Drug Approval for the Treatment of Geographic Atrophy: How We Got Here and Where We Need to Go

Abstract

Purpose:

To discuss the clinical trial results leading to the FDA approval of anti-complement therapies for geographic atrophy (GA), perspectives on functional data from the GA clinical trials, and how lessons from the FDA approval may guide future directions for basic and clinical research in AMD.

Design:

Selected literature review with analysis and perspective

Methods:

We performed a targeted review of publicly available data from the clinical trials of pegcetacoplan and avacincaptad for the treatment of GA as well as scientific literature on the natural history of GA and the genetics and basic science of complement in AMD.

Results:

The approval of pegcetacoplan and avacincaptad was based on an anatomic endpoint of a reduction in the rate of GA expansion over time. However, functional data from two phase 3 clinical trials for each drug demonstrated no visual benefit to patients in the treatment groups. Review of the genetics of AMD and the basic science of the role for complement in AMD provides only modest support for targeting complement as treatment for GA expansion, and alternative molecular targets for GA treatment are therefore discussed. Reasons for the disconnect between anatomic and functional outcomes in the clinical trials of anti-complement therapies are discussed, providing insight to guide the configuration of future clinical studies for GA.

Conclusion:

While avacincaptad and pegcetacoplan are our first FDA-approved treatments for GA, results from the clinical trials failed to show any functional improvement after one and two years, respectively, calling into question whether the drugs represent a “clinically relevant outcome.” To improve the chances of more impactful therapies in the future, we: i) provide basic-science rationale for pursuing non-complement targets, ii) emphasize the importance of ongoing clinical research that more closely pins anatomic features of GA to functional outcomes, and iii) provide suggestions for clinical endpoints for future clinical trials on GA.

Table of Contents Statement:

The FDA approval of anti-complement therapies for geographic atrophy (GA) was based on reducing GA expansion rates, but even after 24 months of treatment, patients demonstrated no benefit in the trial’s pre-specific functional outcomes. We examine the limitations of structural endpoints in GA trials as well as the strength of data supporting the complement hypothesis in GA. We provide suggestions for current GA management, design of future GA trials, and alternate focus areas for GA research.

On February 17, 2023, the FDA announced approval of the anti-complement component 3 (C3) pegylated-peptide, pegcetacoplan (Syfovre, Apellis Pharmaceuticals), for “the treatment of geographic atrophy (GA) secondary to age-related macular degeneration (AMD).”¹ The approval was predicated on the results of two phase 3 clinical trials where pegcetacoplan resulted in a 22% reduction in GA lesion area growth (OAKS trial, which met its primary efficacy endpoint) and a 12% reduction in GA lesion area growth (DERBY trial, which did not meet its primary efficacy endpoint) when given monthly over a 12 month period. Follow-up data at 24 months demonstrated a 22% and 19% reduction in the growth rate of GA lesion area from baseline to 24 months in the two studies when pegcetacoplan was injected monthly.^1,2 The on-going GALE extension study has demonstrated a further reduction in GA growth rate in treated patients from months 24–30.³ In the fall of 2023, the FDA announced approval of an additional anti-complement drug for GA--the pegylated RNA aptamer, avacincaptad pegol, targeting complement factor C5.⁴ Approval for avacincaptad was based upon the results of the GATHER1 and GATHER2 phase 3 clinical trials demonstrating a 27% and 14% respective decrease in growth of GA lesion area with monthly treatment over 12 months.^5,6 Unlike previous approvals of drugs targeting neovascular AMD that used visual acuity as the primary outcome,^7–9 the FDA did not require demonstration of a functional benefit between treated and control arms, such as a difference in visual acuity (VA), scotoma size, or reading speed. In light of the failure of several previous complement modulation trials for GA in AMD,^10,11 and given the modest effects of pegcetacoplan and avacincaptad on GA expansion rates, we explore in this Perspective: 1) how the FDA arrived at GA expansion as a primary study endpoint, 2) the disconnect between GA expansion and VA outcomes in the pegcetacoplan and avacincaptad pegol trials, 3) implications for future GA clinical trial design, and 4) alternative therapeutic pathways outside of the complement pathway that should be tested in future GA trials.

Geographic Atrophy as an Endpoint for Clinical Trials in Age-Related Macular Degeneration

Identifying an achievable and clinically relevant endpoint for GA has been difficult. Because visual acuity loss often occurs only late in the evolution of GA, substantial areas of the macula can be lost before central VA is affected. As a result, using VA as the primary outcome in a GA trial would extend the duration of the study well beyond a realistic time frame for most large-scale randomized clinical trials.¹² Additionally, vision loss in GA patients is highly variable, with previous reports demonstrating a mean decrease in visual acuity of 6.2 letters over 12 months but with a variability of +/− 15.6 letters.¹³ In eyes with 20/80–20/500 vision and foveal GA involvement, visual acuity may actually improve as eccentric preferred retinal loci (PRL) for fixation are discovered and utilized, especially when the fellow eye has a sudden change in visual acuity.¹⁴

GA expansion does, however, lead to dense scotomas over the area of atrophy¹⁵ which can influence activities of daily living, especially reading, independent of central visual acuity. For example, maximum reading rate in AMD patients with GA is highly correlated with the size of the atrophic area, regardless of whether GA is fovea-involving or not.¹⁶

These data were the basis for discussions with the FDA in 2006, leading to the First Open Symposium on Endpoints and Clinical Trial Strategies jointly held between the National Eye Institute (NEI) and the FDA.¹⁷ At that meeting, representatives from both of these federal agencies as well as officials from the Centers for Medicare and Medicaid Services (CMS), university scientists and clinicians, and others discussed, among other topics, how best to evaluate new treatments for AMD.¹⁷ Several presentations at the symposium detailed the impact of expanding GA area on various measures of visual function. Participants also reviewed data demonstrating that both color fundus photography¹⁸ and fundus autofluorescence¹⁹ imaging can be used to reliably quantify the area of absolute scotoma associated with GA lesions. After extensive discussions, the FDA agreed that an acceptable clinical endpoint for GA could be the rate or extent of anatomic progression of GA in treatment and placebo groups. There was an explicit understanding that this difference would represent a clinically relevant outcome even without documentation of a vision function difference. During the clinical endpoints workshop/symposium, the FDA’s Supervisory Medical Director for Ophthalmology specifically stated that “the FDA would consider [a range of different] endpoints for clinical trials, but the trial sponsor would have to justify the clinical relevance of the new endpoints.”¹⁷

Although change in visual acuity is often the primary endpoint in clinical trials for new retinal drug applications, approval of a drug based solely on anatomic outcomes is not unprecedented. For example, trials of the ganciclovir implant for cytomegalovirus (CMV) retinitis did not require visual acuity endpoints, since prior data well demonstrated that expansion of CMV retinitis results in significant vision loss.²⁰ Thus, slowing of the time to retinitis progression from 15 days to 226 days by the ganciclovir implant was viewed as a clinically important outcome.²¹ Similarly, approval of ranibizumab for the treatment of diabetic retinopathy (DR)²² did not require proof of direct visual acuity benefit. Ranibizumab demonstrated a two or three step reduction in diabetic retinopathy severity score (DRSS) of 23% and 14%, respectively, over 24 months.²³ As severity of DRSS correlates with future risk for severe vision loss²⁴, this treatment effect also was believed to represent a clinically relevant outcome.

Discrepancy Between GA Expansion and Functional Outcomes in the Pegcetacoplan Trials

The pegcetacoplan trial results provide an opportunity to test the assumption that slowing GA progression leads to clinically relevant results related to visual function. While the 24 month results from the OAKS and DERBY trials demonstrated an approximately 20% reduction in GA expansion rate in subjects receiving pegcetacoplan, mean visual acuity loss was 7.0, 7.9 and 8.8 letters in the sham, monthly, and every-other-month groups, respectively.²⁵ Given that 31% of patients with GA and good vision demonstrate a 15-letter acuity loss over 24 months²⁶, it was somewhat surprising that neither trial demonstrated any benefit in numerous pre-specified primary or secondary vision/functional outcomes, including the Functional Reading Independence Index and the mean threshold sensitivity on microperimetry. Most surprising was that pegcetacoplan also had no effect on maximum reading speed, a key functional indicator that typically tracks with GA size.²⁵. A similar lack of benefit in pre-specified secondary visual (functional) outcomes was noted in the avacincaptad phase 3 clinical trial results, including standard and low-luminance VA.^5,6

What are the possible explanations for the discrepancy between the reduction in GA growth and lack of any vision function benefit, especially reading speed? Several studies suggest that the location of GA may be as important or more important to reading speed as size of the scotoma.^16,27 For example, scotomas from GA above fixation correlate with faster reading rates while scotomas to the left of fixation correspond to slower reading speeds (for English).

Retrospective analysis of the Chroma and Spectri trials¹⁰ of GA by Chakravarthy and co-authors also found that the relationship between GA and functional outcomes varied by lesion characteristics, including location.²⁸ Lesions with different GA growth rates demonstrated separation in the trajectory of visual acuity loss in subjects with unifocal but not multi-focal lesions. The most significant correlation between GA expansion rates and visual acuity loss was seen in eyes with subfoveal unifocal lesions. A subfoveal, unifocal GA lesion with a growth rate of 1.7 mm²/year corresponded to a 15-letter loss at 96 weeks whereas the same lesion with a GA growth rate of 1.2 mm²/year resulted in a 3-letter loss at 96 weeks. Understanding the effects of pegcetacoplan on similar groups of subjects in the DERBY and OAKS trials will be an important post-hoc analysis that might help guide the ophthalmology community in selecting patients with the greatest potential for therapeutic benefit. Additionally, identifying features in GA patients with good visual acuity that specifically predict vision loss regardless of GA expansion rates will be critical for patient selection in future clinical trials, as recently highlighted by Drs. Chakravarthy and Lad.²⁹

Other explanations for the disconnect between GA expansion and visual acuity outcomes in the pegcetacoplan and avacincaptad trials may stem from how structural endpoints in GA are measured. Fundus autofluorescence (FAF) imaging is the standard metric used by the FDA for assessing GA expansion. However, the relationship between function and altered autofluorescence at the GA border is still not clear.³⁰ Further, pegcetaoplan and avacincaptad appear to increase the risk of choroidal neovascularization that can alter FAF properties and confound GA expansion rate measurements. Finally, blockade of complement may decrease immune-cell mediated removal of hyper-autofluorescent sick and non-functional RPE at the GA border,³¹ creating the appearance of reduced GA expansion but without any functional benefit. Future clinical trials might focus on merging FAF with optical coherence tomography (OCT) imaging to validate GA expansion rates.³². Recent advances in sensitive and accurate GA progression detection using OCT include loss of the ellipsoid zone (EZ) line and hypertransmission defects on en face imaging.^33–35

While the FDA has stated that other structural endpoints, such as photoreceptor preservation, may be utilized in clinical trials for GA,³⁶ structure-function relationships generated from natural history studies³⁷ might not remain constant in retinal tissues exposed to disease modifying therapies. For example, the phase 2 ARCHER study of ANX007, an anti-C1q Fab fragment, demonstrated a significant reduction in multiple assessments of visual acuity loss at 12 months in the treatment groups despite an absence of any statistically significant reduction in GA expansion.³⁸ In contrast, ancillary analysis of OCT data from the phase 2 FILLY trial of pegcetacoplan demonstrated a significant reduction in photoreceptor thinning in patients treated monthly compared with sham subjects^39,40 despite no benefit in pre-specified functional outcomes. Since any structural outcome chosen for a clinical trial is potentially vulnerable to structure-function discrepancy, it would be ideal if drug approval based on a structural outcome alone was conditional on demonstration of a functional benefit in subsequent phase 4 studies.

Re-Examining the Complement Hypothesis in Age-Related Macular Degeneration

Given the modest treatment effect of pegcetacoplan and avacincaptad and prior failures of complement modulators in GA, it may be helpful to re-examine the basic science data supporting the link between complement and AMD. While genome-wide association studies (GWAS) have often failed to yield biologically meaningful results in diseases with complex heritability,⁴¹ the application of GWAS to AMD has consistently yielded evidence for polymorphisms in complement factor H (CFH), a negative regulator of the alternative complement pathway, as being highly correlated with AMD. Polymorphisms associated with other components of the alternative complement pathway have also been found in AMD patients, including CFB, C3, and CFI, although the effect sizes for these other complement genes are significantly smaller than those for CFH.⁴² All of these genetic associations between complement genes and AMD were discovered primarily by comparing high-risk intermediate or advanced AMD patients to control populations. Interestingly, while CFH and other alternative complement pathway genes have been shown in numerous subsequent studies to be important for the onset of advanced AMD,^43,44 they have failed to show a significant correlation with progression of advanced AMD (namely, GA expansion rates). For example, analysis of the AREDS2 cohort showed no relationship between CFH alleles and GA expansion rates and, somewhat paradoxically, an inverse relationship between the C3 AMD risk alleles and GA expansion. These results were confirmed in additional studies of the AREDS and other cohorts.^45,46 The AREDS1 cohort also failed to demonstrate a link between the most common CFH risk allele and GA expansion.⁴⁷ A more recent GWAS study spanning multiple patient cohorts meticulously separated out patients with new onset of GA from those with expansion of existing GA; this study also failed to find any link between the complement system and GA expansion rates.⁴⁸ The authors are aware of only two studies that have shown a potential link between CFH and GA progression rates. One study linked GA progression with the most common CFH risk allele based on a small cohort of 26 patients, with no statistical analysis.⁴⁹. The other study contrasted the association of CFH SNPs (single nucleotide polymorphisms) with prevalence of GA vs. GA expansion rates in a cohort of approximately 300 patients in Spain.⁵⁰ While CFH SNPs were associated with GA prevalence at p values <0.001, p values for association of CFH SNPs with GA progression rates were 0.04. Overall, the genetic evidence for complement pathway involvement in GA progression is weak and contradictory.

Faced with strong genetic evidence between CFH and GA onset, there was understandable enthusiasm for clinical trials of anti-complement agents after the initial AMD GWAS studies were published in 2005. However, concurrent with a rise in interest in anti-complement therapy for AMD, the joint FDA/NEI Symposium on clinical endpoints had developed a consensus that the rate of GA expansion should be utilized as a primary endpoint for dry AMD trials. The flurry of anti-complement trials that ensued therefore used GA expansion rates to determine primary efficacy rather than onset of GA. Some of these trials were small, including C5 inhibition (administered systemically as eculizimab in a phase 2 trial involving 30 patients,¹¹ administered intravitreally as tesidolumab in a phase 2 trial involving 158 patients)⁵¹ and properdin inhibition (intravitreal CLG561, phase 2 trial of 75 patients).⁵² Others were large, including CFD inhibition (intravitreal lampalizumab, two phase 3 trials with nearly 1900 combined patients),¹⁰ CFI overexpression (intraocular gene therapy GT005, two phase 2 trials with over 350 combined patients; discontinued by Novartis on September 11, 2023), and C3 inhibition (intravitreal NGM621, phase 2 trial involving 320 patients).⁵³ Unfortunately, all these trials, and a few others not listed, failed, but they provided us with the needed data to surmise that there is a disconnect between the complement hypothesis and GA expansion.

Why might complement inhibition have such modest effects on GA expansion? While many detrimental effects of the core convergent components of the complement system, C3 and C5, have been shown in animal and cell culture models of retinal degeneration,^54–57 there is also evidence that both components have important roles in retinal homeostasis.⁵⁸ The complement system may mark debris for clearance, helping remove apoptotic cells and reduce inflammation. Further, low-level signaling via receptors for the activated form of C3 and C5 may actually promote neuronal survival. Indeed, knockouts in mice of C3 or the receptors for activated C3 or C5 result in retinal degeneration.^58,59 Thus, any potential beneficial effects of C3 or C5 inhibition may be partly offset by loss of homeostatic functions for these factors. The good/bad ratio of complement inhibition may be retinal layer specific. For example, while there may be larger deleterious effects of complement inhibition in the neural retina, the beneficial effects of complement inhibition may be prominent in the RPE or choroid.

As CFH variants represent the overwhelming genetic risk factor for onset of advanced AMD among all complement components,⁶⁰ non-complement functions of CFH may also be driving AMD pathogenesis. Indeed, the major AMD risk allele in CFH is in a site of the protein that does not affect complement activation but does affect the trapping of lipids in Bruch’s membrane.⁶¹ Data therefore suggests that CFH may modulate AMD risk by altering Bruch’s membrane to promote drusen formation rather than altering C3 or C5 activation (the targets of pegcetacoplan and avacincaptad).^62–64 The other major risk allele for sporadic advanced AMD, ARMS2/HTRA1^{65 66}, is thought to involve a protein that also remodels Bruch’s membrane.⁶⁷ Further, all three of the Mendelian disorders that mimic sporadic AMD, namely Sorsby’s Fundus Dystrophy,⁶⁸ Doyne Honeycomb Dystrophy,^69,70 and Late-Onset Retinal Dystrophy,⁷¹ all involve proteins that are also involved in Bruch’s membrane remodeling. Thus, the major risk alleles and Mendelian genes for AMD or AMD-like processes appear to converge on Bruch’s membrane alterations rather than the complement pathway, arguing that other factors besides complement may be primary drivers of dry AMD, including GA.

Alternatives to Complement Inhibition for Future Geographic Atrophy Trials

Given the present modest effects of anti-C3 and anti-C5 inhibition on atrophic AMD, alternative therapeutic pathways are being considered. As GA is fundamentally a cell death phenomenon, therapies that directly target cell death pathways may hold promise in slowing GA expansion. Blockade of the transmembrane protein Fas not only stops caspase-mediated cell death signaling cascades but also dampens production of proinflammatory cytokines and chemokines, which may otherwise trigger the death of adjacent cells.⁷² Another therapeutic approach may involve targeting ARMS2/HTRA1. Compared with CFH, the ARMS2/HTRA1 risk allele is more consistently associated with GA progression.⁶⁵ As HTRA1 is a secreted protease that modifies the extracellular matrix, modulation of HTRA1 may improve diffusion across Bruch’s membrane and mitigate choroidal hypoperfusion that may contribute to GA expansion. While HTRA1 inhibitors are in clinical development, it is currently unclear whether HTRA1 activity needs to be enhanced or suppressed to alleviate AMD pathology.^66,73 A comprehensive analysis of RPE tissue from AMD donors with the HTRA1 risk allele demonstrated lower expression of the protein,⁶⁶ suggesting that augmentation rather than inhibition of HTRA1 expression may be needed for AMD patients. In contrast, HTRA1 overexpression in animal models induces AMD-like changes to Bruch’s membrane.^67,74 Modulating HTRA1 to precisely the correct activity for benefit in AMD may prove difficult. Indeed, Roche terminated its phase 2 study of an antibody targeting HTRA1 in late 2022.⁷⁵

Other strategies for targeting GA include modulating mitochondrial dysfunction, a well-documented pathology in all stages of AMD.⁷⁶ Alterations in mitochondrial function may directly impact activation of complement⁷⁷, such that mitochondria-directed therapies may simultaneously target two pathways with links to AMD. Additionally, the role of hypoxia, manifested by the loss of choriocapillaris and reductions in choroidal perfusion,^78,79 is being probed as a key factor for AMD disease progression. Factors that work to blunt the toxic byproducts of the visual cycle, that are known to collect in drusen, are also in development. Finally, strategies to modulate inflammation outside of complement regulation may be useful in GA. For example, connexins mediate inflammatory signaling from stressed cells to neighboring cells, and downregulating connexins may help stop the “inflammatory spread” of GA.⁸⁰ Alternatively, siglec receptors are involved in immune regulation and could be modulated to prevent GA expansion.⁸¹

A Path Forward

With the approval of pegceptacoplan and avacincaptad which for now is limited to the United States, the ophthalmic community will need to determine whether the modest reduction in the rate of GA expansion ultimately leads to a consistent visual function benefit for patients. This information is even more needed given the recent negative opinion by the European Medicines Agency (EMA) on approval for pegcetacoplan.⁸² The rationale for this rejection was “Although the studies showed that Syfovre slowed the growth of geographic atrophy lesions, this did not lead to clinically meaningful benefits for patients. It was noted that benefits of a treatment should impact patients’ everyday functioning, and this was not demonstrated in the studies.”

Given that industry cannot perform every study prior to drug approval, we believe the following investigations may be helpful in that effort:

1. Assess on-going post-hoc analyses of the DERBY, OAKS, GATHER1, and GATHER2 data for their utility in hypothesis generation and selection of possible primary endpoints for future clinical trials, including improvements in microperimetry test results.

2. Develop large post-marketing clinical trials. One of many possible trials would be to interrogate eventual functional benefits of GA therapies approved strictly with a structural outcome endpoint. The DRCR Retina Network is well-positioned to develop such studies and a number of protocol ideas are already under consideration. The GALE extension study³ for pegcetacoplan will report 36-month results in September 2025, and this should provide longer follow-up data to determine whether slowing the rate of GA growth leads to functional changes specifically in normal and low luminance BCVA.

3. Measure GA expansion in future GA trials by multimodal analysis, at minimum with FAF accompanied by OCT, including assessment of EZ loss expansion or quantification of choroidal hyper-transmission defects by en face imaging.

4. In future clinical trials, the location and structure of the GA lesion should be carefully considered and paired with appropriate functional outcomes (e.g., GA to the right of fixation with a corresponding scotoma to the left). Suggestions by Lad and Chakravarthy²⁹ to enrich future GA clinical trial cohorts with patients who have unifocal, subfoveal lesions may be considered, as these patients appear to have the best correlation between GA expansion rates and visual acuity decline. In addition, it is likely that artificial intelligence analysis of patient and imaging characteristics will help enrich clinical trial cohorts with those most likely to functionally respond to treatment.

5. Given that GWAS studies have not shown a relationship between the complement system and expansion of GA, complement modulation trials could evaluate the progression of intermediate AMD to advanced AMD, where the GWAS data is much stronger. Appropriate endpoints may include determining progression of incomplete RPE and outer retinal atrophy (iRORA) to complete RPE and outer retinal atrophy (cRORA)⁸³ or examining the development and progression of hypertransmission choroidal defects³⁵ in parallel with appropriate demonstrations of functional benefits.

6. Genotyping of patients entering clinical trials may help determine how well HTRA1, CFH, and other common polymorphisms predict response to therapy.^43,44

7. Alternative therapeutic strategies for GA outside of complement inhibition could be more robustly investigated and supported.

Although additional investigations guided by such recommendations will eventually provide further clarity about the role for complement inhibition in the management of GA, at present patients and their physicians face treatment decisions based on today’s level of understanding. In counseling our patients about the risks and benefits of pegcetacoplan therapy, we have converged on the following set of talking points, which may be helpful for other providers: 1) The therapy has no proven benefit for visual acuity, reading speed, or other measures of visual function, and it is unclear how long one must treat beyond two years for such a benefit to emerge. 2) Treatment with monthly or bimonthly injections for 2 years can be expected to delay a given amount of GA expansion, on average, by 4.5 months.^1,2,25 The ongoing GALE study will help determine the magnitude of the delay in GA expansion associated with longer treatment periods. 3) Unlike in wet AMD, where treatment burden and cost typically diminish over time, patients can expect to undergo indefinite therapy every 4–8 weeks, in many cases in both eyes, as we lack imaging biomarkers that allow us to individualize the dosing regimen. 4) While all intravitreal therapies carry risk, the combined side-effect profile for pegcetacoplan is more significant than that of the most commonly-used agents for wet AMD and includes ischemic optic neuropathy, occlusive retinal vasculitis, and induction of exudative AMD.^1,2,25,84 Given the data from the GATHER1 and GATHER2 studies, our conversation with patients on avacincaptad will be broadly similar to our conversation about pegcetacoplan.

Pegcetacoplan, avacincaptad, and other drugs that follow may be welcome options for some of our dry AMD patients with GA. Our hope is that information from real-world, post-approval publications will help refine the types of measurement tools and endpoint data needed for future GA clinical trials and, most importantly, lead to therapies with more substantial clinically relevant outcomes for patients.