2026 年 8 巻 1 号 p. 28-35
The Fight Retinal Blindness! Registry is a prospectively-designed registry developed in 2009 that collects international data on treatment outcomes for eye diseases including neovascular age-related macular degeneration, diabetic macular edema and retinal vein occlusion. The validated and high-quality data have generated significant real-world evidence regarding clinically relevant issues, such as long-term visual outcomes, dosing regimens and practitioner variation. This paper outlines the Fight Retinal Blindness! Registry, highlighting its pros and cons, and introduces some findings from the registry.
Clinical registries are collections of information on individuals with specific diseases to provide a structured foundation for improving healthcare outcomes1),2). The field of ophthalmology has increasingly used clinical registries to monitor care quality, benchmark practitioners’ performances, investigate practice variations, track outcomes of treatments and procedures and document adverse events1),2). This paper introduces the Fight Retinal Blindness! (FRB!) registry, an international registry of treatment outcomes of retinal and other eye diseases.
The FRB! registry was developed in 2009 to track and analyse real-world patient treatment outcomes for neovascular age-related macular degeneration (nAMD) with its subsequent expansion to other retinal diseases such as other courses of macular neovascularization, diabetic macular edema and retinal vein occlusion3–5). The registry is a prospectively-designed, efficient, web-based data collection tool to monitor treatment outcomes for retinal diseases in routine clinical practice. Any practitioner with access to the Internet can join the FRB! registry from a range of devices such as Windows-PC, Macintosh, tablet computers and smartphones. The web system allows practitioners to review their own data and benchmark their outcomes against national benchmarks at any time5). Since 2022 users with electronic medical records (EMR) have been able to extract the information from their EMR, convert it into a specific format (comma-separated values file) and upload it onto the FRB! registry through single point data entry. This means that the information regarding a particular patient is only collected once during routine clinical practice in the EMR and is then subsequently transferred to the FRB! database. Since 2025 this step has been expanded using the Fast Health Interoperability Resources (https://fhir.org/) interface for automatic connections and data transfers between a practitioner’s EMR and the FRB! registry server6). The FRB! registry has expanded to include modules that track outcomes of treatment in other fields in ophthalmology such as glaucoma (Fight Glaucoma Blindness!), cornea (Fight Cornea Blindness!), uveitis (Fight Uveitis Blindness!), uveal melanoma (Fight Tumour Blindness!), retinopathy of prematurity (Fight Childhood blindness!) and inherited retinal diseases (Fight Inherited Retinal Blindness!). These are collectively called Save Sight Registries, of which the FRB! registry is the flagship module.
As of April 2025, the nAMD, diabetic macular edema and retinal vein occlusion modules in the FRB! registry had 155, 72 and 56 active practitioners (i.e. had entered in the last 12 months) with 6332, 1292 and 753 eyes actively tracked (i.e. that had had any data entered in the last 6 months), respectively. The practices participating in the FRB! registry are spread internationally in countries such as Australia, Switzerland, Netherlands, Spain, France, Lebanon and Japan (Fig. 1).
Each black dot represents a practice. Only the practices that had entered any data in the last one year were included.
The FRB! project is composed of the Steering Committee, the Publishing Committee and the User groups5). The Steering Committee supervises the development of the project and coordinates various requests from the User groups. The members of the Steering Committee are retinal experts elected among the active users. The Publishing Committee oversees proposals for analysis submitted by primary investigators to check if the topic is clinically important and the study design and statistical methods to be used are adequate. The primary investigators can perform analysis using data of other users only 1) if the Publishing Committee approves the proposal and 2) if the users do not oppose the use of their data which is established for each user for each analysis. The biostatisticians, located in Australia and Switzerland, undertake the analysis for international projects. For local projects, i.e. in a particular country, local analysts can analyse their own national data. Here again, users are asked if they agree for their data to be used for a particular analysis. The User groups are the practitioners who enter the data into the registry and are responsible for the development of new modules or the modification of existing modules.
Data FieldsPractitioners or their assistants retrospectively enter data for baseline visits (visits at treatment initiation) and follow-up visits, having agreed to track ≥85% of eyes with the relevant condition that they treat. Treatment decisions such as drug choice and regimen are made at the practitioner’s discretion in consultation with the patient, reflecting routine clinical practice. The burden of data entry by retinal practitioners, who are busy with their routine clinical practice, is reduced by the adoption of a prospectively identified, minimum dataset that is mandatory at each visit. This aligns with the International Consortium of Healthcare Outcomes Measures standard outcome set for macular degeneration7). The data fields for the nAMD module are listed in Table 1.
| Field | Pre-specified options | Mandatory |
|---|---|---|
| —Baseline visit— | ||
| Gender | Female, male or undisclosed | Yes |
| Year of birth | Year of birth | Yes |
| Ethnicity | White/Caucasian, Asian, Hispanic, etc. | Yes |
| Smoking status | Smoker, Ex-smoker, Non-smoker or unknown/undisclosed | No |
| Angiographic lesion type | Occult, Minimally classic, Predominantly classic, Retinal angiomatous proliferation, Disciform scar, idiopathic polypoidal choroidal vasculopathy, Juxtapapillary or not done | Yes |
| Greatest linear dimension | The longest diameter (mm) across a lesion measured by fluorescein angiography | No |
| Previous treatment | VEGF inhibitors (e.g. aflibercept 2 mg, ranibizumab, etc.), steroid (e.g. dexamethasone, fluocinolone, etc.), photodynamic therapy, cataract surgery, etc. | Yes |
| Ocular conditions | Early AMD, Dry AMD, Axial myopia, vitrectomy, etc. | Yes |
| —Follow-up visit— | ||
| Treatment | VEGF inhibitors (e.g. aflibercept 2 mg, ranibizumab, etc.), steroid (e.g. dexamethasone, fluocinolone, etc.), photodynamic therapy, etc. | Yes |
| Visual acuity | Number of letters read on a logarithm of the minimum angle of resolution scale. Best of uncorrected, corrected or pinhole. May be entered in Snellen | Yes |
| Intraocular pressure | Intraocular pressure in mmHg | No |
| Choroidal neovascularisation activity | Inactive, active with subretinal fluid only, or active with any combination of fluid excluding subretinal fluid only | Yes |
| Macular atrophy | Center-affecting atrophy, non center-affecting atrophy or no atrophy | Yes |
| Subretinal fibrosis | Center-affecting fibrosis, non center-affecting fibrosis or no fibrosis | Yes |
| Ocular adverse events | Anterior uveitis, Chorioretinitis, occlusive retinal vasculitis, non-occlusive retinal vasculitis, vitritis, infectious endophthalmitis, non-infectious endophthalmitis, retinal pigment epithelium tear, etc. | Yes |
| Ocular events | Other procedures given not directly related to treatment for nAMD such as cataract extraction or vitrectomy | Yes |
| Discontinuation | Reason for discontinuation of patients from database, including non-treatment related reasons such as death | No |
AMD = age-related macular degeneration, VEGF = vascular endothelial growth factor.
Firstly, the data in the FRB! registry are validated and high-quality. Users can finalise data only when all the mandatory fields are entered, which reduces the amount of missing data. All categorical data fields (e.g. ethnicity) have pre-specified options (e.g. White/Caucasian, Asian, etc.) so that users cannot enter free text. All continuous data fields (e.g. visual acuity) have pre-specified ranges (e.g. 0–100 logarithm of the minimum angle of resolution letters) so that users cannot enter abnormal values. In a previous validation study the data in the FRB registry were checked against the health records for around 250 randomly sampled eyes from 11 practitioners, showing good accuracy (>95%) for the fields such as sex, birth year, treatment and visual acuity (VA) (+/− 5 letters)8).
Secondly, the FRB! registry helps practitioners understand their own practice patterns and results. Practitioners can view their patients’ treatment journeys using concise and interactive graphs. Hovering over the graph enables the practitioner to view additional details at each visit such as time and change in visual acuity from the index visit and the previous visit. They can also see the summary of their own treatment outcomes (e.g. average visual acuity gain for the eyes they treated by year) and compare them with the benchmark in the registry using a funnel plot (Fig. 2). These web-based features inform practitioners’ treatment decisions and help patients understand their disease and treatment compliance, all enhancing the quality of care and treatment outcomes.
Each dot represents one practitioner with the two curves of 95% control limits. Practitioners with better, normal and worse performance fall above the upper curve, within the two curves and below the lower curve, respectively. The x-axis represents the expected number of eyes adjusted by age, gender, baseline visual acuity, an initial type of vascular endothelial growth factor inhibitors, fluid type at baseline (subretinal fluid-only or intraretinal fluid) and the existence of subretinal fibrosis and macular atrophy. The y-axis represents the difference between the observed and expected values. The central dot lines represent the reference standard.
Firstly, practitioners or their assistants who still use paper-based files or where they are unable to extract the collected information from the EMR, have to enter data via a web interface separately from the patient’s health record. This additional burden may prevent some data from being entered. Technological advances and the increasing adoption by practices worldwide of EMRs now provide options for single point data entry to replace dual-point data entry through the FRB! web-interface through either comma-separated value-based data transfer or the Fast Health Interoperability Resources. Single point entry of observational data significantly reduces the data entry burden for practitioners, making the FRB! registry more attractive to practitioners and its data more representative of the general population.
Secondly, practitioners cannot follow patients who switch to another practitioner who is not part of the FRB! users network. These issues likely contribute to the 10% missing visits that are estimated to occur in the registry’s database compared to the health records8).
nAMD, characterized by macular neovascularization which leads to subretinal and intraretinal fluid and hemorrhage, is a leading cause of vision loss in older adults9). The first-line treatment is intravitreal injections of vascular endothelial growth factor (VEGF) inhibitors, which can prevent VA loss and often improve VA by suppressing macular neovascularization activity10),11). Practitioners administer injections monthly for the first 1–3 injections in an induction phase until the lesion becomes inactive before moving to a maintenance phase where injections are given less frequently12). Treatment regimens in routine clinical practice are broadly classified as reactive or proactive based on the timing of dosing in the maintenance phase4). The reactive regimen requires monthly visits with treatments when the lesion becomes active, while proactive regimens deliver treatments at every visit irrespective of macular neovascularization activity but gradually extend the injection intervals after macular neovascularization has been inactivated to keep the lesion inactive with as few visits as possible, which helps with patient compliance10).
An FRB! analysis investigated 24-month outcomes for treatment-naïve eyes with nAMD that received ranibizumab from practitioners who stated they used proactive regimens10). The 1198 eligible eyes identified experienced a mean VA gain of +5.3 letters with a mean of 13 injections and 14.8 visits over two years. Another FRB! analysis similarly investigated 24-month outcomes for aflibercept 2 mg, a drug that was released around 6 years after ranibizumab11). The 136 eligible eyes achieved a mean VA gain of +6.0 letters with a mean of 13.5 injections and 15.2 visits. These studies found that the eyes treated with proactive regimens attended approximately 40% fewer visits than if they had been monitored monthly (i.e. 24 visits) with a good visual outcome at 24 months.
A more recent FRB! analysis directly compared 24-month outcomes of proactive and reactive regimens within the FRB! registry for treatment-naïve nAMD eyes starting treatment with aflibercept 2 mg or ranibizumab13). The 2070 eyes treated by a proactive regimen had a significantly greater mean improvement in VA (+4.2 letters) than the 1243 eyes treated by a reactive regimen (+1.3 letters) with more injections (mean, 15 vs. 10) and a similar number of visits (mean, 15.8 vs. 16.3) over 24 months, providing evidence for the superiority of proactive vs. reactive treatment in real-world practice. Interestingly, the reduction in visits that proactive regimens usually deliver was not evident, likely because the patients treated with a reactive regimen were not followed as closely, or at least did not attend as often, as is recommended. Practitioners need to plan an optimal treatment regimen for every patient where good visual outcomes and financial, physical and mental burden caused by more injections are well balanced.
Long-term OutcomesThe FRB! registry provides high-quality long-term data. Many studies have reported 1- and 2-year outcomes for eyes with nAMD treated by VEGF inhibitor injections in real-world practice, but data on ≥5-year outcomes are still limited. An FRB! analysis in 2015 evaluated the outcomes of 1212 eyes with nAMD, of which 45% (549/1212) had completed 5 years of follow-up12). They found that VA improved by 5.3 letters after 6 months and remained above the baseline VA for ≥5 years with a median of ≥5 injections given annually throughout the follow-up period. These findings were better than the limited other data available at that time – a post hoc analysis of a landmark clinical trial reported that VA deteriorated by 8.6 letters at around 7 years with a much lower frequency of injections14). Another post-hoc analysis reported VA loss at 5 years with fewer injections than the FRB! study15). Practitioners participating in the FRB! registry, most of whom were Australian, were more likely to have used a proactive regimen than the other studies where patients might have been under-treated.
An FRB analysis in 2020 reported 10-year outcomes for eyes with nAMD for the first time16). They found that the eyes from Australia and New Zealand had a mean visual change of only –0.9 letters with a median of 53 injections over ten years, while the eyes from Switzerland lost a mean of 14.9 letters with a median of 42 injections. Another FRB analysis evaluated 10-year outcomes in France, where the eyes had a mean visual loss of 18 letters with a median of 27.5 injections. Other studies investigating 10-year outcomes for eyes with nAMD are listed in Table 2. It is consistently found that more frequent injections are associated with better long-term visual outcomes. Using the values in Table 2, we performed a linear model where VA change at 10 years was regressed on the number of injections over 10 years and baseline VA with the number of eyes for each study used as frequency weights. A significant association was observed between the number of injections and VA change over 10 years (+0.63 letters per 1-injection increase; P = 0.002).
| Study | Single/Multi-centre | Country | No. of eyes |
Baseline VA, letters |
VA change over 10 years |
Injections |
|---|---|---|---|---|---|---|
| Gillies M et al. 2020 | Multi-centre (FRB!) | Australia/New Zealand | 132 | 60.7 | −0.9 | 53 (median) |
| Gillies M et al. 2020 | Multi-centre (FRB!) | Switzerland | 37 | 61.6 | −14.9 | 42 (median) |
| Wolff B et al. 2020 | Multi-centre (FRB!) | France | 116 | 57.5 | −18.0 | 27.5 (median) |
| Chandra S et al. 2020 | Single-centre | UK | 149 | 59.5 | −2.1 | 52.2 |
| Starr MR et al. 2020 | Single-centre | US | 130 | 54.5 | −13.0 | 45.1 |
| Spooner K et al. 2021 | Single-centre | Australia | 293 | 60 (median) | 3.0 | 58.1 |
| Hujanen P et al. 2023 | Single-centre | Finland | 43 | 61 | −4.3 | 53.8 |
| Brynskov T et al. 2020 | Single-centre | Denmark | 74 | 61.5 | −5.0 | 42.9 |
VA = visual acuity.
Values represent mean or n unless ‘median’ is written.
Australia may have better results than many other countries probably because VEGF inhibitors are relatively unrestricted and reimbursed by the Australian government and there are a sufficient number of ophthalmologists in Australia who are paid adequately per injection to perform them. Practitioners should balance the better visual outcomes that come with more injections with the physical and financial burden that comes with them, along with the ever-present risk of intraocular infection and inflammation that comes with every injection. Long-term data in the FRB! registry, which will extend even longer moving forward, will be increasingly valuable since nAMD needs to be treated continuously until the patients’ death.
Outlier AnalysesA logical progression of collecting and analysing treatment outcome data involves benchmarking clinical performance among practitioners to further enhance visual outcomes17). The inter-practitioner analyses may facilitate peer learning, whereby practitioners with suboptimal outcomes are identified and supported to improve their practice through feedback and comparison with peer benchmarks17). Additionally, the use of registry data could help to identify systemic barriers to care delivery at the patient level, such as financial constraints and geographic remoteness with limited access to transportation18),19). These processes will contribute to a self-improving healthcare system, ultimately reducing patient harm and enhancing treatment outcomes17).
An FRB! analysis compared the practices of individual practitioners (38 practitioners, 1266 eyes) in Australia based on their treatment performance for their patients with nAMD20). The practitioners were classified into a high outlier (n = 1), non-outliers (n = 34) and low outliers (n = 3) using a funnel plot with the indicator of mean adjusted change in VA during the first 12 months of treatment. The eyes treated by the high outlier, non-outliers and low outliers had crude mean VA changes of 7.5, 5.1 and 2.5 letters, respectively. They found that the high outlier gave significantly more injections of VEGF inhibitors within 12 months than the non-outliers (mean, 8.6 vs. 7.7) followed by the low outliers with significantly fewer injections (mean, 7.1). In keeping with this, the high outlier had the shortest 12-month injection interval, followed by the non-outliers and low outliers (9.4 weeks, 10.8 weeks and 11.8 weeks, respectively).
Another FRB! analysis compared the 24-month outcomes of eyes with retinal vein occlusion managed by individual practitioners (29 practitioners, 1100 eyes) in multiple countries based on their frequency of treatment using VEGF inhibitors21). Three outlying “intensive” practitioners, 22 “typical” practitioners and 4 outlying “relaxed” practitioners were identified using a funnel plot with the indicator of crude mean injection interval. They found that the eyes treated by the intensive practitioners had the best improvement in mean adjusted change in visual acuity, followed by the typical and relaxed practitioners (+16.2, +13.6 and +9.3 letters, respectively).
The two FRB! studies have indicated that some practitioners might have provided suboptimal treatment, resulting in poorer patient outcomes20),21). Some patients have difficulty receiving injections at the recommended frequency due to financial limitations or other barriers, such as physical disabilities, mental health issues or living in remote areas with limited access to ophthalmic care18),19). On the practitioner side, injection frequency might have been restricted by concerns about a higher risk of developing macular atrophy and endophthalmitis, high patient caseloads or low expectations regarding VA improvement18),19). Practitioners classified as low outliers may have been unaware that their patients’ outcomes were significantly worse than those of their peers, underscoring the need for them to be informed of this. Providing this information may help practitioners adopt more effective treatment regimens and identify targeted interventions for patients such as subsidised transport services.
While some jurisdictions have established mandatory outcome monitoring systems, participation in the FRB! project is voluntary17). Thus, feedback must be provided in a fully anonymised manner to encourage practitioners to continue their participation. Implementing a confidential notification process to inform ophthalmologists with outlying results about their performance is another way that the FRB! system can facilitate improvements in patient visual outcomes and overall quality of care.
Validation StudyThe FRB! registry is a database that was prospectively designed for specified eye diseases (e.g. nAMD). This is an advantage over administrative claims databases where diseases are identified by inference from International Classification of Diseases-10 codes and the use of drugs and procedures related to that disease. However, since the FRB registry is not automatically synchronized with medical records, some errors might occur when transcribing data from medical records to the registry and some visits might be missed.
The FRB! investigators performed a validation study by checking data recorded in the FRB! registry against medical records using randomly sampled 245 eyes from 11 practitioners8). Gender, birth year, previous treatments received, treatments given at each visit and VA were included in the outcomes. An error was regarded as a discrepancy between the medical records and the data entered in the FRB! registry. Errors in VA were regarded as ≥5-letter absolute difference between the two values. Missing visits were defined as visits present in the medical records that were not entered in the FRB! database.
They found that VA had an error rate of 5.1% and that of the other fields was <5%, showing that all fields had good to excellent agreement. The overall error rate was 3.5%. The positive predictive value and negative predictive value for visits were 99% and 89%, suggesting that visits entered in the FRB! registry are almost perfectly reliable, while absent visits in the FRB! registry have a little chance of actually having visits in the medical records. The proportion of missing visits was 10.2%. This study found quite high accuracy (>95%) of the FRB! registry although some visits might be missed, likely due to the high burden of entering data. Results of FRB! analyses should be interpreted with this in mind, particularly for studies that focus on visit gaps and treatment gaps.
The FRB! project has provided a significant amount of real-world evidence on intravitreal injections for macular disease using its high-fidelity data, which has helped improve treatment outcomes in ophthalmology. It should be acknowledged that the registry depends absolutely on the selfless contribution of many practitioners. Several new drugs for nAMD, such as brolucizumab, faricimab and aflibercept 8 mg, the so-called second-generation VEGF inhibitors, have been released in recent years22–24). Biosimilars for the traditional agents have started to appear on the market to reduce financial burden for patients and healthcare systems25). Tyrosine kinase inhibitors are another promising drug that might reduce the treatment burden as they are eye drops26). The FRB! registry will continue to analyse and present essential real-world evidence for these new drugs as well as the traditional drugs.
Professors Gillies and Barthelmes are inventors of the software used to collect the data for this analysis.
The Fight Retinal Blindness! Project is supported by a grant from unrestricted educational grants from Bayer, Roche and National Health and Medical Research Council (#1195021). Those from Bayer and Roche have no grant numbers because they are funds from pharmaceutical companies rather than from governments.
None.
Research Design: YH, DB and MG
Data Acquisition and/or Research Execution: YH, DB and MG
Manuscript Preparation: YH, DB and MG
YH is the guarantor.
