T he scope and sophistication of clinical trials in the field of retina have expanded dramatically in the last decade, encompassing novel pharmacologic, biologic, and gene-based interventions across a range of conditions causing vision loss. Recently, the U.S. Food and Drug Administration (FDA) approvals of avacincaptad pegol (Izervay) and pegcetacoplan (Syfovre) in 2023 represented the first approved treatments for dry age-related macular degeneration (dAMD) and geographic atrophy (GA), respectively. ^, For neovascular AMD (nAMD) and diabetic macular edema (DME), an abundance of available anti-VEGF therapies—including ranibizumab (Lucentis), aflibercept (Eylea), bevacizumab (Avastin), faricimab (Vabysmo) and corresponding biosimilars (Cimerli, PavBlu)—has transformed management by improving efficacy and accessibility. ^,,, Despite this progress in the past few years, the results of clinical trials are frequently neutral or modest, raising critical questions about how study design, execution and feasibility factors can shape the success of retinal research. Understanding these aspects is essential for translating innovation and potential into meaningful clinical advancements and for refining the foundation of evidence that guides patient care. ^,, Sustained investments into clinical research is vital to developing and generating evidence for alternative and innovative treatments for disease, while making them more accessible for usage by addition to the public market. Previous analyses have focused primarily on isolated examples of clinical trials or disease-specific difficulties, causing limited insight into the characteristics of interventional studies that have been terminated early, withdrawn, or completed without achieving their primary endpoints and have not included the significance of their phases to their probability of success. ^,, This review systematically examines a decade of registry data encompassing terminated, withdrawn, completed-negative, and FDA rejected retinal trials to identify common elements of trial design that may enhance the success of future investigations.

Although the causes of trial failure are multifactorial, several recurring challenges are particularly prominent within clinical research in retina. This includes the strict procedures that are associated with the study protocols and the inconsistencies that can vary across different locations when it comes to study procedures being conducted, including application of the inclusion and exclusion criteria. ^, Inclusion and exclusion criteria that is provided for the study also may also create a lack of diversity and leading to an increase of miscategorization of patient’s eligibility and disease severity The heavy reliance on imaging-based inclusion criteria—such as strict thresholds for central subfield thickness (CST) or lesion morphology on optical coherence tomography (OCT)—can substantially limit participant eligibility and slow recruitment. ^, Along with these imaging requirements, the reading centers, as well as any additional vendors that are outsourced to determine specific eligibility requirements related to the trial, such as clinical labs, CROs, and site management organization, can provide inconsistent determinations and results, leading to more variable results upon completion of the trial. Ensuring that a clinical trial has ample resources and a thorough team to review and interpret data that is collected for a clinical trial in an efficient manner, with ample feedback whenever applicable. ^,, Even at the site level, there is a possibility of investigator bias and treatment variability for trials that have supplemental or rescue treatments if the investigative product is not sufficient in addressing the indication, which can skew results that are presented to the sponsor. ^,

Similarly, demanding visit schedules, complex injection regimens, and the need for precise endpoint measurement contribute to operational strain and participant attrition. Emerging therapies, including gene and cell-based approaches, introduce additional feasibility barriers related to small target populations, specialized surgical techniques, and extended follow-up requirements that require more staff members and additional time for preparation and follow-ups. These factors collectively underscore how the unique scientific and logistical landscape of retinal trials magnifies the risk of noncompletion or negative efficacy outcomes. Despite increasing recognition of these challenges, few studies have attempted to systematically evaluate unsuccessful retinal clinical trials as a group. Existing discussions tend to emphasize either specific disease categories, such as nAMD or DME, or focus narrowly on specific terminated studies alone. ^, As a result, there remains limited understanding of the broader design- and execution-related patterns that distinguish successful programs from those that fail to achieve efficacy or regulatory approval. By synthesizing data across a decade of publicly registered interventional retina trials in their most pivotal stages, this review seeks to identify recurring design features and operational pitfalls that contribute to trial nonsuccess and a evaluate limitations for clinical trial submissions for FDA approval and review.

The purpose of this analysis is to provide a comprehensive assessment of unsuccessful retinal clinical trials and to extract practical insights for the design and feasibility evaluation of future studies. By identifying modifiable contributors in the analysis of pivotal efficacy and confirmatory stages prior to regulatory submission (phases 2-3)—such as restrictive eligibility criteria, insufficient sample sizes, location/dispersion of sites, or overall high procedural complexity—this work aims to inform strategies that enhance trial efficiency, minimize research waste, prioritize patient treatment timelines, increase patient satisfaction, and strengthen the translational pathway from innovation to clinical application.

This work was designed as a retrospective, registry-based review of interventional retinal clinical trials at their most impactful stages, during phases 2 and 3. The objective was to identify and characterize factors associated with unsuccessful study outcomes—defined as termination before completion, withdrawal before enrollment, completion without efficacy, or lack of regulatory approval. The analysis focused on interventional trials registered over a ten-year period (September 1st, 2015 to September 1st, 2025) and included only industry-sponsored investigations looking to submit their clinical trial data to the FDA in phases 2 and 3. The focus on phase 2 and 3 trials for this review comes from their importance to gather evidence to confirm safety and efficacy of the investigative product for humans to proceed with submission to the FDA and subsequent testing. These trials contain vital data that is significant to FDA approval of investigative products for public and professional usage.

Trial data were extracted from ClinicalTrials.gov using the Aggregate Analysis of ClinicalTrials.gov (AACT) database. Search terms included “geographic atrophy,” “neovascular age-related macular degeneration,” “diabetic retinopathy,” “dry age-related macular degeneration,” “diabetic macular edema,” and other related terms, including phase number. This was done to gather information for clinical trials targeting these specific conditions in the field of retina and gain better understanding of the distribution of trials by condition and phase. The trials that completed the duration of their data collection or were terminated were sorted according to the time frame of completion between September 1st, 2015 and September 1st, 2025. Duplicate listings and observational studies were excluded, as well as trials that were exclusively conducted outside of the United States. The interventional trials that were included in the analysis based on the criteria above were then sorted accordingly to observe any trends or patterns that may lead to trial nonsuccess. The clinical trials were grouped by similar indications, as well as by phase, intervention type, and main reason for nonsuccess.

A total number of 58 interventional clinical trials were identified in the registry review spanning September 1st, 2025 to September 1st, 2025 as completed or terminated. After applying eligibility filters to include phase 2 and 3 industry-sponsored studies that encompasses pharmacologic, gene-based, and biologic therapies targeting major retinal disorders, including GA, dAMD, nAMD, DME, and DR, a total of 26 trials met the inclusion criteria for analysis. Of these trials, a substantial portion were terminated early, withdrawn, or completed without achieving primary efficacy endpoints, while a much smaller subset reached FDA submission. Trial-level data was compiled by study phase, therapeutic class, and trial outcome to identify design, feasibility, and operational factors that were observed with nonsuccess. Indications that were related or associated were grouped together to simplify the categories of clinical trials. A complete list of the individual trials included in this analysis, with detailed information on sponsor, intervention, mechanism of action, phase, and outcome, is provided in Supplemental Table S1 .

Table 1 summarizes all phase 2 and 3 interventional clinical trials in retina that were completed in the United States between 2015 and 2025, categorized by disease indication and trial phase. A total of 58 interventional studies were identified across all indications that were focused on, as mentioned earlier. The highest overall attrition was observed in GA/dAMD trials (52.4%), followed by nAMD trials (34.6%) and DME/DR trials (23.1%). Phase 2 studies showed proportionally greater failure overall compared to phase 3.

Following the phase-specific and indication-specific analysis, additional patterns were observed when considering the nature of the investigational products and interventions themselves. As retinal drug development has diversified and innovation has continued, variation in the invention type and their mechanism of action has become increasingly prominent across trials. There has been a movement from transitional anti-VEGF and corticosteroid agents towards more complex and experimental modalities–including gene and cell-based therapies and sustained-release delivery systems–which while mechanistically innovative, may introduced increased procedural complexity and feasibility constraints. The distribution of these therapeutic approaches and their failure rates are summarized in Table 2 , which outlines the relative contribution of each intervention and mechanism class to overall nonsuccess of clinical trials.

Across all the unsuccessful trials that were analyzed, there were several recurrent themes that emerged. The most common reasons that were discovered were missed primary efficacy endpoints, futility determinations at interim analyses, and strategic sponsor decisions. Operational challenges, including high procedural burden and potential data variability related to complex imaging requirements and site-level inconsistencies, were frequently observed across trials. In the earlier stages of clinical trials, like phase 2, there was overall a higher rate of trials not being successful, with the main reasoning for failure being related to not being able to achieve the primary efficacy endpoint.

As summarized in Table 3 , differences were observed across trial phases in the distribution of unsuccessful outcomes. Phase 2 trials more frequently included discontinuations following interim analyses or failure to meet predefined endpoints, whereas phase 3 trials were more commonly characterized by failure to achieve primary efficacy endpoints at study completion. A comprehensive list of included trials and detailed study characteristics are provided in Supplemental Table S1 (available at AJO.com ). Additionally, an overview of these trials and their study timelines and key assessments is provided in Supplemental Table S2 (also available at AJO.com ). This may provide additional insight into the operational constraints at the site level, as well as feasibility factors from the sponsor-level to the patient-level.

Across the decade of clinical trials analyzed, the retinal research landscape reflected meaningful scientific advancement but was also characterized by persistently high rates of trial non-completion and neutral outcomes. Among the 58 U.S. phase 2 and phase 3 interventional studies identified, 44.8% failed to meet their primary endpoints or were terminated early, with over two-thirds of these failures occurring during phase 2. This distribution highlights how early-stage feasibility and efficacy challenges may constrain the successful transition to pivotal development. Higher attrition was observed in nAMD and GA/dAMD trials, when compared to DR and DME programs. When considered alongside differences in investigational approaches, additional patterns were noted across trials. Trials involving more complex or novel delivery approaches may have greater operational and procedural demands compared to those assessing established IVT anti-VEGF therapies. In addition to the possibility of increased demands, these programs involve distinct safety considerations that warrant further investigation in the context of study participation and protocol adherence.

Many of these unsuccessful efforts shared features of elevated procedural complexity, which may be associated with data variability, compromise the accuracy of imaging and functional endpoints, and increase the likelihood of protocol deviations. High-frequency visits, intricate imaging protocols, and demanding surgical or injection procedures may introduce inconsistencies in data capture and magnify the risk of missing or poor-quality data, due to the overwhelming number of assessments and data points that need to be captured. These operational pressures not only affect the reliability of efficacy assessments but may also obscure genuine therapeutic effects, leading to inconclusive or misleading outcomes despite otherwise promising biological rationales. The assessment of an investigational product’s safety profile in a trial typically necessitates more frequent study visits to monitor for possible complications or unexpected reactions, allowing sponsors to gather longitudinal safety data and confirm tolerability following treatment exposure through observation.

Retinal clinical trials with high procedural complexity—such as those involving invasive delivery, frequent imaging or injection visits, or gene/cell therapy protocols—may face increased operational demands and risk of non-completion. As of lately, the retinal development landscape has been characterized by increasing trial complexity, including advanced delivery systems with more stringent eligibility and imaging requirements. Intervention types and mechanisms of action were examined across trials to explore potential patterns within these categories. With each indication in retina, there is demand for certain types of interventions, including oral pill medications for dAMD and gene therapies being investigated for the exudative conditions of nAMD and DME, which are less invasive compared to their currently FDA approved counterparts. With more invasive gene therapy or stem cell trials that involve surgical interventions, the more frequent safety visits can be surrounding the surgery date, as well as more intensive and assessment-heavy visits with more imaging and vision testing to ensure patient safety while collecting pivotal data points. With IVT treatments focusing on anti-VEGF mechanisms and the complement pathways, there was an increase of failed clinical trials in these categories compared to all of the others. The complement system is a known component to the progression of dAMD, as well as GA, and is a pathway that plays a central role in the innate immune response. Dysregulation of complement activation contributes to chronic inflammation, formation of drusen and progressive retinal pigment epithelium damage— processes that make the complement components appealing yet complex therapeutic targets.

This increase in failures may be due to the complicated nature of the complement system and the dual action anti-VEGF mechanisms. The complement system has been implicated in the development of AMD and GA, with six members of the complement cascade (CFB, CFH, CFI, C2, C3 and C9) accounting for 40 to 60 percent of AMD heritability, which differ significantly in concentration corresponding with stages and act as inflammatory foci. This has positioned complement components as key therapeutic targets, despite challenges in selectively targeting individual factors. Age-related macular degeneration (AMD) is a retinal disease involving the dysfunction of macular photoreceptors, RPE, Bruch membrane, and choroid, leading to a loss of central vision and features like macular drusen, RPE atrophy, and CNV depending on the severity. Geographic atrophy (GA) represents an advanced form of AMD characterized by larger scale RPE destruction in the macular region and photoreceptor loss. In addition to this, the biological heterogeneity and genetic variability in complement factor alleles associated with AMD may partly explain why many complement-targeted therapies have struggled to demonstrate clinically significant results, since there is patient-specific genetic variability. The heterogeneity in enrolled patient populations, including variation in disease stage, lesion characteristics, and prior treatment exposure, may also contribute to variability in outcomes and reduce the ability to detect consistent treatment effects.

Behind the failed clinical trials, many sponsors and pharmaceutical companies responsible for the trials had immediate financial consequences to manage. Due to the unsuccessful results of these trials, the companies were often found in poor financial circumstances and had layoffs as a result. Oftentimes, sponsors and pharmaceutical companies undergo restructuring due to focuses of project management and alignment of company goals, which can lead to subsequent repercussions for other initiated projects and established contracts in action. This result leads to an exhaust in resources at sites and vendors alike, which could have been avoided with better planning and thorough reviews of data by the sponsor team during ongoing data collection. The main example included in this analysis was the OXEYE trial that was created by Oxular Limited for DR, which was terminated due to a “non-safety related sponsor-decision to stop further recruitment.” With this trial, there was only a total number of 3 patients enrolled before enrollment was terminated, however, there were also 3 patients total experiencing “systemic treatment emergent AEs” per the reported data. The sponsor reasoning for termination was very vague, but it must be emphasized that the trial was open for enrollment since prior to October 2023, which was when the first patient in the trial was dosed. Following this, the trial was announced that it would be closing October 2025. Other trials that were impacted by changes in business strategy were sponsored by Ionis Pharmaceuticals and Cognition Therapeutics for dAMD/GA, although the details behind these trial closures were more vague and discrete, as the companies stated that recruitment and enrollment did not begin for the trial and that the trial was closed as a part of a strategic business decision, respectively. Both trials mentioned did not have any results posted. ^,

Beyond scientific and operational contributors to trial non-success, the downstream effects on sponsoring organizations themselves emerged as an important theme in this analysis. Many of the failed or discontinued programs were observed with significant corporate consequences, including restructuring, shifts in strategic direction, and changes in leadership. These organizational responses often reflect the financial sensitivity of ophthalmic biotechnology companies, which commonly rely on a small number of lead programs to support continued development. As a result, pivotal trial outcomes—whether neutral, negative, or halted for futility—can trigger rapid and substantial changes in company structure, workforce composition, or executive oversight. The following examples illustrate how unsuccessful late-stage retinal trials have directly influenced the stability and trajectories of several sponsors included in this review. For the Opthea COAST and ShORe trials, their non-success resulted in dramatic structure changes in the company, rather than the company’s business decisions influencing the progression of the clinical trial itself, with the possibility that the company could be required to pay its investors more than one billion dollars due to the previous Development Funding Agreements (DFA). In addition to owing its investors, Opthea also had undergone restructuring of the company after the failure of their phase 3 trials and discontinuation of their product development, resulting in over 80% layoffs and a significant reduction in its board of directors by over 50% as well. ^, In contrast, Adverum previously announced in 2022 that it had cut 38% of jobs within their company to focus on their development of their gene therapy for nAMD and redirect costs to make progress on the trial, which was included in this analysis. Similar consequences resulted downstream from trial failures were observed from other trials as well, specifically GLEAM and GLIMMER that were initiated by Kodiak Sciences. Following the failure of these trials to meet their respective primary endpoints, the CMO and chief development officer of Kodiak Sciences had resigned from their position, highlighting how pivotal clinical outcomes can directly shape organizational stability within the pharmaceutical industry.

In DME and DR clinical trials, where investigational therapies often strive to enhance durability and lower treatment-related risks, these aims introduce their own set of design and feasibility challenges. Across Ocuterra therapeutics and Kodiak sciences studies’ difficulties of reaching their primary endpoint and Adverum’s safety-driven issues, these examples illustrate the need to not only consider the systemic pressures that shape trial performance, but the significance of biological mechanisms and how novel pathways can contribute to variability across indications. ^,,, With Novartis Pharmaceutical’s study it was determined that their investigational product had around the same efficacy as the comparator in the study, Lucentis, so the development of the drug did not continue following completion of the trial. This further emphasizes the importance of feasibility and efficacy of medications. Differences across indications were observed, with comparatively lower attrition in DME and DR trials relative to nAMD and GA programs. This may reflect differences in disease pathophysiology, endpoint responsiveness, and the maturity of existing therapeutic options across indications.

To further investigate a possible relationship between interventions and mechanisms to probability of success or positive outcomes, Table 2 was created to break down the distribution of the analyzed trials into separate categories. The main categories of intervention that were observed to have high rates of failure or attrition were IVT injections focusing on dual pathway or combination inhibition (20%) and complement pathway modulators (29%). Since many components of the complement pathway can contribute to the phenotype of dAMD and eventually GA, it is unsurprising that clinical trials with interventions that focus on complement pathway modulation have the highest percentage of non-success out of all categories. The multi-factorial nature of dAMD can make it a difficult indication to target with therapies, as patients will have genetic variability. More invasive or procedure-intensive approaches, including gene therapy–based interventions, were examined to assess patterns of attrition across therapeutic categories. These trials often involved greater assessment burden and more complex data collection requirements, which may represent potential sources of variability. In this analysis, gene therapy trials were observed within a mid-range distribution of unsuccessful outcomes. The extent to which procedural burden, biological variability, or therapeutic effect contributes to these patterns cannot be determined and warrants further investigation. However, these trials are often observed with more complex study visits and involve rescue or supplemental treatment if the investigational therapy is not maintaining the patient at baseline disease stability. The use of rescue criteria in retina trials may further complicate endpoint interpretation, as early intervention with standard therapies can attenuate differences between treatment arms, with variability introduced by investigator-dependent decision-making.

Variation across intravitreal intervention strategies was also observed. Anti-VEGF monotherapy and combination or dual-pathway approaches were represented across unsuccessful trials at varying frequencies, although the influence of mechanistic complexity or incremental therapeutic benefit remains unclear. Independently, anti-VEGF drugs can reduce the increase of vascular permeability caused by the increase of VEGF, inhibit the growth of CNV, alleviate ME and increase the visual of patients that are diagnosed with nAMD, but these are often injected repeatedly on a regular (typically monthly) basis, which can increase the risk of ocular complications. In addition to this, the high efficacy benchmark established by existing anti-VEGF therapies also creates a challenging landscape for investigational monotherapies and combination therapies, as this was a common primary endpoint observed. When used in conjunction with other drugs administered intravitreally, the risks associated can compound and may be accompanied by serious adverse reactions. As summarized in Table 3 , reasons for unsuccessful outcomes were distributed across several categories, including lack of efficacy, interim futility assessments, and safety-related findings. Differences were observed between trial phases, with phase 2 studies more frequently discontinued following interim analyses and phase 3 studies more often failing to meet primary efficacy endpoints. This pattern likely reflects the differing objectives of exploratory and confirmatory trial design. In retinal therapies, this may be further influenced by high efficacy benchmarks, such as BCVA-based endpoints, which may limit sensitivity in detecting incremental benefit, particularly in patients with relatively preserved baseline vision. Ceiling effects may also contribute, as patients with higher baseline acuity have less room for measurable improvement.

While understanding the causes of a trial failure is essential, these patterns gain additional meaning when considered within the broader context of the clinical development pathway. Early-phase success does not consistently translate into late-phase outcomes, particularly as trials expand in scale, duration, and geographic distribution. The main examples that were discovered in this analysis were Opthea Therapeutics’ COAST and ShORe trials, which were briefly touched on earlier, as well as Outlook Therapeutics’ NORSE EIGHT trial. The COAST and ShORe trials were preceded by Opthea’s phase 2b trial, which demonstrated that combination therapy with sonizinibercept reduced the proportion of patients experiencing vision loss by 82%. The sponsor also stated they observed up to a 42% relative increase in participants with 20/40 vision at Week 24 of the study. The phase 2b trial focused on the investigational drug’s impact on different lesion types with the comparator drug being ranibizumab and had a total of 366 participants across 109 sites. This marked a stark difference from the organizational approach for the phase 3 trials, which combined had 1984 participants enrolled across a total of 400 sites with comparator drugs involved in the investigation were both aflibercept and ranibizumab. ^,,, Since the discontinuation of the clinical trial came following site initiations and enrollment of participants, it is possible that the broader focus on participants and comparator drugs had led to a detriment due to their prior success in a more uniform and smaller sample size. Larger trials are multi-center, have broader patient demographics, in addition to physician practices, which are less regulated and more variable in clinical practice. The timelines of the clinical trials are different as well, as phase 2b evaluated participants for 24 weeks, whereas the phase 3 trials observed them for 52 weeks and greater for long term follow-up. ^,,, Generally speaking, phase 3 trials often use stricter efficacy criteria and have longer follow-up, which can reveal that the added benefit of the combination therapy over standard anti-VEGF is not clinically meaningful.

As for Outlook Therapeutic’s NORSE EIGHT trial, the data was successfully reviewed and submitted to the FDA for approval but subsequently declined by the agency in August 2025. The FDA’s declination marks another setback for the company, after it had withdrawn its application for the drug previously in 2022 after they had requested additional information. The company had also previously submitted an application August 2023, which was rejected due to “several CMC issues, open observations from pre-approval manufacturing inspections, and a lack of substantial evidence.” In an announcement made by Outlook Therapeutics, they stated that the FDA issued a complete response letter (CRL) to its biologics license application (BLA) for solely a lack of substantial evidence of effectiveness due to ONS-5010 not meeting the primary efficacy endpoint in the NORSE EIGHT trial. In the prior trial, named NORSE TWO, primary endpoints for effectiveness were met and it was recommended that confirmatory evidence be submitted to support the company’s biologics application. While Outlook Therapeutics works with the FDA to ensure that the requirements are met for their BLA, they plan on continuing their efforts to expand into additional markets in Europe, as ONS-5010 has been commercially available in Germany and the UK for the treatment of nAMD since June 2025. ^, The progression of the NORSE trials started with NORSE ONE, which took place at 9 sites in Australia and involved 61 participants. The next NORSE trial was NORSE TWO, a phase 3 pivotal trial that took place at 39 sites in the U.S. with 228 patients. Subsequently, the NORSE THREE trial was a phase 4 open-label safety study in the U.S. with 197 participants, while NORSE EIGHT was a phase 3 trial that took place in the U.S. across 61 sites and 400 participants. ^, Together, these patterns illustrate how early-phase successes do not always translate into late-phase reliability, particularly when studies scale rapidly in sample size, geographic dispersion, and operational complexity. This may reflect increased heterogeneity in patient populations, site-level practices and endpoint variability in large, multi-center trials. These examples together emphasize the importance of ensuring biological plausibility but also operational feasibility when transitioning from exploratory trials to pivotal development.

Although the findings of this analysis provide valuable insight into the patterns and factors behind trial non-success, several important limitations should be considered when interpreting these results. Firstly, this analysis focused on publicly available information from ClinicalTrials.gov , as well as sponsor-provided information or statements. These sources vary in completeness and level of detail, which can limit the depth of comparative analysis. This may also introduce some bias towards publicly available information, which can potentially underrepresent more negative or neutral outcomes that remain unreported by the sponsors, therefore, lacking true transparency. Similarly, there were limitations with considering other factors that may have impacted trial operations, such as mergers, financial burdens, and their success in other global markets. Secondly, this review focuses on U.S. phase 2 and phase 3 interventional trials. Although this aligns with the aim of the review to examine pivotal efficacy and confirmatory stages, it excludes earlier phase safety trials which may display different failure patterns and feasibility constraints. While clinical trials can have a variety of locations located within the United States and internationally, it is important to note that each regulatory region operates under distinct standards for drug approval and market authorization. For therapies seeking FDA approval in the U.S., the success and integrity of the pivotal phases of the clinical trial are critical, as these studies must meet the agency’s expectations for safety, efficacy, and methodological rigor. Thirdly, the analysis uses descriptions and categorizations of failure patterns rather than statistical inference to predict the outcome of clinical trial success. Even though this type of analysis is appropriate for the variety of trials and the purpose of the dataset, the analysis itself does not establish direct causal relationships and cannot determine which design features definitively increase trial failure risk. Accordingly, the findings of this analysis should be interpreted as hypothesis-generating rather than indicative of causal relationships.

The results of this study can provide direction to future clinical trials and investigators alike, as being aware of existing patterns can help prevent them and allow others to learn from them. Being that this analysis focused on the most influential stages of clinical trials, the goal was to determine the common patterns amongst unsuccessful interventional retina trials in the United States. By examining various elements such as intervention type, mechanism of action, procedural complexity, and operational demands placed on sites, this work seeks to clarify which design choices and feasibility constraints most strongly influence whether a therapy progresses through pivotal development, to FDA submission, and hopefully approval. Understanding these patterns is essential, not only for contextualizing past failures, but also for informing the planning, optimization, and execution of clinical trials across the retina space moving forward. Across phase 2 and 3 studies, nearly 45% of trials were unsuccessful, with the highest percentages of nonsuccess observed in nAMD and dAMD/GA programs, mainly concentrated during evaluation of early efficacy. These patterns underscore the importance of anticipating indication-specific challenges–such as the lesion growth in GA and the CST thickness values in nAMD–when selecting endpoints, determining target populations, and predicting realistic timelines for demonstrating therapeutic benefits. Future clinical trials can benefit from more feasibility modeling during protocol development, especially when expanding from more controlled phase 2 populations to multi-center, heterogenous phase 3 populations. The distribution of failure reasons in this data set demonstrated that nonsuccess frequently arises from the combination of biological and logistical factors. Bringing attention to these elements early in trial development may help mitigate risks that become amplified as studies expand and progress on a global scale. In addition to this, strengthening site training, implementing early data-quality monitoring, and standardizing data points whenever possible may reduce the variability that has contributed to inconclusive or poor outcomes in a number of the programs analyzed. Intervention-specific trends identified from this analysis revealed that gene therapies, complement inhibitors, and dual-action inhibitors tended to be observed with higher rates of nonsuccess, which may be driven by procedural requirements or insufficient efficacy signals. Conversely, failures of oral or topical ophthalmic agents suggest issues regarding lack of efficacy, reinforcing the need for robust pharmacodynamic data and clearer biologic rationale prior to investing in later-stage development. As trial design becomes increasingly sophisticated, emerging analytical tools, including those derived from artificial intelligence, may offer meaningful support. Machine learning models applied to historical trial datasets may help identify design features associated with trial nonsuccess and could provide early signals of potential feasibility challenges. Additionally, AI-supported standardization of imaging assessments has the potential to reduce site-level inconsistency, improving the interpretability of anatomical endpoints. By integrating evidence-based feasibility considerations, modality-specific design principles and thorough endpoint planning into trial development, future interventional trials may be better equipped to achieve success and deliver innovative therapeutic options for patients with vision-threatening conditions and diseases. Moving forward, greater emphasis on aligning early-phase signals with realistic late-stage expectations, including endpoint sensitivity and operational feasibility, may be critical to improving translational success in retinal drug development. As the therapeutic landscape in retina continues to expand, applying lessons learned from unsuccessful programs will be essential for reducing development insufficiencies, minimizing patient and resource burden, and accelerating the advancement of clinically meaningful treatments.