NIHR Moorfields Biomedical Research Centre

Central serous chorioretinopathy (CSC) is a common cause of central vision loss, primarily affecting men 20-60 years of age. To date, no consensus has been reached regarding the classification of CSC, and a wide variety of interventions have been proposed, reflecting the controversy associated with treating this disease. The recent publication of appropriately powered randomised controlled trials such as the PLACE trial, as well as large retrospective, non-randomised treatment studies regarding the treatment of CSC suggest the feasibility of a more evidence-based approach when considering treatment options. The aim of this review is to provide a comprehensive overview of the current rationale and evidence with respect to the variety of interventions available for treating CSC, including pharmacology, laser treatment, and photodynamic therapy. In addition, we describe the complexity of CSC, the challenges associated with treating CSC, and currently ongoing studies. Many treatment strategies such as photodynamic therapy using verteporfin, oral mineralocorticoid antagonists, and micropulse laser treatment have been reported as being effective. Currently, however, the available evidence suggests that half-dose (or half-fluence) photodynamic therapy should be the treatment of choice in chronic CSC, whereas observation may be the preferred approach in acute CSC. Nevertheless, exceptions can be considered based upon patient-specific characteristics.

BACKGROUND: Radical measures are required to identify and reduce blindness due to diabetes to achieve the Sustainable Development Goals by 2030. Therefore, we evaluated the accuracy of an artificial intelligence (AI) model using deep learning in a population-based diabetic retinopathy screening programme in Zambia, a lower-middle-income country. METHODS: We adopted an ensemble AI model consisting of a combination of two convolutional neural networks (an adapted VGGNet architecture and a residual neural network architecture) for classifying retinal colour fundus images. We trained our model on 76 370 retinal fundus images from 13 099 patients with diabetes who had participated in the Singapore Integrated Diabetic Retinopathy Program, between 2010 and 2013, which has been published previously. In this clinical validation study, we included all patients with a diagnosis of diabetes that attended a mobile screening unit in five urban centres in the Copperbelt province of Zambia from Feb 1 to June 31, 2012. In our model, referable diabetic retinopathy was defined as moderate non-proliferative diabetic retinopathy or worse, diabetic macular oedema, and ungradable images. Vision-threatening diabetic retinopathy comprised severe non-proliferative and proliferative diabetic retinopathy. We calculated the area under the curve (AUC), sensitivity, and specificity for referable diabetic retinopathy, and sensitivities of vision-threatening diabetic retinopathy and diabetic macular oedema compared with the grading by retinal specialists. We did a multivariate analysis for systemic risk factors and referable diabetic retinopathy between AI and human graders. FINDINGS: A total of 4504 retinal fundus images from 3093 eyes of 1574 Zambians with diabetes were prospectively recruited. Referable diabetic retinopathy was found in 697 (22·5%) eyes, vision-threatening diabetic retinopathy in 171 (5·5%) eyes, and diabetic macular oedema in 249 (8·1%) eyes. The AUC of the AI system for referable diabetic retinopathy was 0·973 (95% CI 0·969-0·978), with corresponding sensitivity of 92·25% (90·10-94·12) and specificity of 89·04% (87·85-90·28). Vision-threatening diabetic retinopathy sensitivity was 99·42% (99·15-99·68) and diabetic macular oedema sensitivity was 97·19% (96·61-97·77). The AI model and human graders showed similar outcomes in referable diabetic retinopathy prevalence detection and systemic risk factors associations. Both the AI model and human graders identified longer duration of diabetes, higher level of glycated haemoglobin, and increased systolic blood pressure as risk factors associated with referable diabetic retinopathy. INTERPRETATION: An AI system shows clinically acceptable performance in detecting referable diabetic retinopathy, vision-threatening diabetic retinopathy, and diabetic macular oedema in population-based diabetic retinopathy screening. This shows the potential application and adoption of such AI technology in an under-resourced African population to reduce the incidence of preventable blindness, even when the model is trained in a different population. FUNDING: National Medical Research Council Health Service Research Grant, Large Collaborative Grant, Ministry of Health, Singapore; the SingHealth Foundation; and the Tanoto Foundation.

PURPOSE: The Argus II Retinal Prosthesis System (Second Sight Medical Products, Inc, Sylmar, CA) was developed to restore some vision to patients blind as a result of retinitis pigmentosa (RP) or outer retinal degeneration. A clinical trial was initiated in 2006 to study the long-term safety and efficacy of the Argus II System in patients with bare or no light perception resulting from end-stage RP. DESIGN: Prospective, multicenter, single-arm clinical trial. Within-patient controls included the nonimplanted fellow eye and patients' native residual vision compared with their vision with the Argus II. PARTICIPANTS: Thirty participants in 10 centers in the United States and Europe. METHODS: The worse-seeing eye of blind patients was implanted with the Argus II. Patients wore glasses mounted with a small camera and a video processor that converted images into stimulation patterns sent to the electrode array on the retina. MAIN OUTCOME MEASURES: The primary outcome measures were safety (the number, seriousness, and relatedness of adverse events) and visual function, as measured by 3 computer-based, objective tests. Secondary measures included functional vision performance on objectively scored real-world tasks. RESULTS: Twenty-four of 30 patients remained implanted with functioning Argus II Systems at 5 years after implantation. Only 1 additional serious adverse event was experienced after the 3-year time point. Patients performed significantly better with the Argus II on than off on all visual function tests and functional vision tasks. CONCLUSIONS: The 5-year results of the Argus II trial support the long-term safety profile and benefit of the Argus II System for patients blind as a result of RP. The Argus II is the first and only retinal implant to have market approval in the European Economic Area, the United States, and Canada.

PURPOSE: Transplantation of human embryonic stem cell (hESC)-derived retinal pigment epithelial (RPE) cells offers the potential for benefit in macular degeneration. Previous trials have reported improved visual acuity (VA), but lacked detailed analysis of retinal structure and function in the treated area. DESIGN: Phase 1/2 open-label dose-escalation trial to evaluate safety and potential efficacy (clinicaltrials.gov identifier, NCT01469832). PARTICIPANTS: Twelve participants with advanced Stargardt disease (STGD1), the most common cause of macular degeneration in children and young adults. METHODS: Subretinal transplantation of up to 200 000 hESC-derived RPE cells with systemic immunosuppressive therapy for 13 weeks. MAIN OUTCOME MEASURES: The primary end points were the safety and tolerability of hESC-derived RPE cell administration. We also investigated evidence of the survival of transplanted cells and measured retinal structure and function using microperimetry and spectral-domain OCT. RESULTS: Focal areas of subretinal hyperpigmentation developed in all participants in a dose-dependent manner in the recipient retina and persisted after withdrawal of systemic immunosuppression. We found no evidence of uncontrolled proliferation or inflammatory responses. Borderline improvements in best-corrected VA in 4 participants either were unsustained or were matched by a similar improvement in the untreated contralateral eye. Microperimetry demonstrated no evidence of benefit at 12 months in the 12 participants. In one instance at the highest dose, localized retinal thinning and reduced sensitivity in the area of hyperpigmentation suggested the potential for harm. Participant-reported quality of life using the 25-item National Eye Institute Visual Function Questionnaire indicated no significant change. CONCLUSIONS: Subretinal hyperpigmentation is consistent with the survival of viable transplanted hESC-derived RPE cells, but may reflect released pigment in their absence. The findings demonstrate the value of detailed analysis of spatial correlation of retinal structure and function in determining with appropriate sensitivity the impact of cell transplantation and suggest that intervention in early stage of disease should be approached with caution. Given the slow rate of progressive degeneration at this advanced stage of disease, any protection against further deterioration may be evident only after a more extended period of observation.

PURPOSE: Retinitis pigmentosa (RP) is a group of inherited retinal degenerations leading to blindness due to photoreceptor loss. Retinitis pigmentosa is a rare disease, affecting only approximately 100 000 people in the United States. There is no cure and no approved medical therapy to slow or reverse RP. The purpose of this clinical trial was to evaluate the safety, reliability, and benefit of the Argus II Retinal Prosthesis System (Second Sight Medical Products, Inc, Sylmar, CA) in restoring some visual function to subjects completely blind from RP. We report clinical trial results at 1 and 3 years after implantation. DESIGN: The study is a multicenter, single-arm, prospective clinical trial. PARTICIPANTS: There were 30 subjects in 10 centers in the United States and Europe. Subjects served as their own controls, that is, implanted eye versus fellow eye, and system on versus system off (native residual vision). METHODS: The Argus II System was implanted on and in a single eye (typically the worse-seeing eye) of blind subjects. Subjects wore glasses mounted with a small camera and a video processor that converted images into stimulation patterns sent to the electrode array on the retina. MAIN OUTCOME MEASURES: The primary outcome measures were safety (the number, seriousness, and relatedness of adverse events) and visual function, as measured by 3 computer-based, objective tests. RESULTS: A total of 29 of 30 subjects had functioning Argus II Systems implants 3 years after implantation. Eleven subjects experienced a total of 23 serious device- or surgery-related adverse events. All were treated with standard ophthalmic care. As a group, subjects performed significantly better with the system on than off on all visual function tests and functional vision assessments. CONCLUSIONS: The 3-year results of the Argus II trial support the long-term safety profile and benefit of the Argus II System for patients blind from RP. Earlier results from this trial were used to gain approval of the Argus II by the Food and Drug Administration and a CE mark in Europe. The Argus II System is the first and only retinal implant to have both approvals.

IL-23 and Th17 cells are key players in tissue immunosurveillance and are implicated in human immune-mediated diseases. Genome-wide association studies have shown that the IL23R R381Q gene variant protects against psoriasis, Crohn's disease and ankylosing spondylitis. We investigated the immunological consequences of the protective IL23R R381Q gene variant in healthy donors. The IL23R R381Q gene variant had no major effect on Th17 cell differentiation as the frequency of circulating Th17 cells was similar in carriers of the IL23R protective (A) and common (G) allele. Accordingly, Th17 cells generated from A and G donors produced similar amounts of Th17 cytokines. However, IL-23-mediated Th17 cell effector function was impaired, as Th17 cells from A allele carriers had significantly reduced IL-23-induced IL-17A production and STAT3 phosphorylation compared to G allele carriers. Our functional analysis of a human disease-associated gene variant demonstrates that IL23R R381Q exerts its protective effects through selective attenuation of IL-23-induced Th17 cell effector function without interfering with Th17 differentiation, and highlights its importance in the protection against IL-23-induced tissue pathologies.

BACKGROUND: In chronic central serous chorioretinopathy (CSCR), fluid accumulates in the subretinal space. CSCR is a common visually disabling condition that develops in individuals up to 60 years of age, and there is no definitive treatment. Previous research suggests the mineralocorticoid receptor antagonist, eplerenone, is effective for treating CSCR; however, this drug is not licensed for the treatment of patients with CSCR. We aimed to evaluate whether eplerenone was superior to placebo in terms of improving visual acuity in patients with chronic CSCR. METHODS: This randomised, double-blind, parallel-group, multicentre placebo-controlled trial was done at 22 hospitals in the UK. Participants were eligible if they were aged 18-60 years and had had treatment-naive CSCR for 4 months or more. Patients were randomly assigned (1:1) to either the eplerenone or the placebo group by a trial statistician through a password-protected system online. Allocation was stratified by best-corrected visual acuity (BCVA) and hospital. Patients were given either oral eplerenone (25 mg/day for 1 week, increasing to 50 mg/day for up to 12 months) plus usual care or placebo plus usual care for up to 12 months. All participants, care teams, outcome assessors, pharmacists, and members of the trial management group were masked to the treatment allocation. The primary outcome was BCVA, measured as letters read, at 12 months. All outcomes apart from safety were analysed on a modified intention-to-treat basis (participants who withdrew consent without contributing a post-randomisation BCVA measurement were excluded from the primary analysis population and from most secondary analysis populations). The trial is registered with ISRCTN, ISRCTN92746680, and is completed. FINDINGS: Between Jan 11, 2017, and Feb 22, 2018, we enrolled and randomly assigned 114 patients to receive either eplerenone (n=57) or placebo (n=57). Three participants in the placebo group withdrew consent without contributing a post-randomisation BCVA measurement and were excluded from the primary outcome analysis population. All patients from the eplerenone group and 54 patients from the placebo group were included in the primary outcome. Modelled mean BCVA at 12 months was 79·5 letters (SD 4·5) in the placebo group and 80·4 letters (4·6) in the eplerenone group, with an adjusted estimated mean difference of 1·73 letters (95% CI -1·12 to 4·57; p=0·24) at 12 months. Hyperkalaemia occurred in eight (14%) patients in each group. No serious adverse events were reported in the eplerenone group and three unrelated serious adverse events were reported in the placebo group (myocardial infarction [anticipated], diverticulitis [unanticipated], and metabolic surgery [unanticipated]). INTERPRETATION: Eplerenone was not superior to placebo for improving BCVA in people with chronic CSCR after 12 months of treatment. Ophthalmologists who currently prescribe eplerenone for CSCR should discontinue this practice. FUNDING: Efficacy and Mechanism Evaluation Programme, and National Institute for Health Research and Social Care.

Primary diabetes care and diabetic retinopathy (DR) screening persist as major public health challenges due to a shortage of trained primary care physicians (PCPs), particularly in low-resource settings. Here, to bridge the gaps, we developed an integrated image-language system (DeepDR-LLM), combining a large language model (LLM module) and image-based deep learning (DeepDR-Transformer), to provide individualized diabetes management recommendations to PCPs. In a retrospective evaluation, the LLM module demonstrated comparable performance to PCPs and endocrinology residents when tested in English and outperformed PCPs and had comparable performance to endocrinology residents in Chinese. For identifying referable DR, the average PCP's accuracy was 81.0% unassisted and 92.3% assisted by DeepDR-Transformer. Furthermore, we performed a single-center real-world prospective study, deploying DeepDR-LLM. We compared diabetes management adherence of patients under the unassisted PCP arm (n = 397) with those under the PCP+DeepDR-LLM arm (n = 372). Patients with newly diagnosed diabetes in the PCP+DeepDR-LLM arm showed better self-management behaviors throughout follow-up (P < 0.05). For patients with referral DR, those in the PCP+DeepDR-LLM arm were more likely to adhere to DR referrals (P < 0.01). Additionally, DeepDR-LLM deployment improved the quality and empathy level of management recommendations. Given its multifaceted performance, DeepDR-LLM holds promise as a digital solution for enhancing primary diabetes care and DR screening.

The occurrence of Graft-versus-Host Disease (GvHD) is a prevalent and potentially lethal complication that develops following hematopoietic stem cell transplantation. Humanized mouse models of xenogeneic-GvHD based upon immunodeficient strains injected with human peripheral blood mononuclear cells (PBMC; "Hu-PBMC mice") are important tools to study human immune function in vivo. The recent introduction of targeted deletions at the interleukin-2 common gamma chain (IL-2Rγ(null)), notably the NOD-scid IL-2Rγ(null) (NSG) and BALB/c-Rag2(null) IL-2Rγ(null) (BRG) mice, has led to improved human cell engraftment. Despite their widespread use, a comprehensive characterisation of engraftment and GvHD development in the Hu-PBMC NSG and BRG models has never been performed in parallel. We compared engrafted human lymphocyte populations in the peripheral blood, spleens, lymph nodes and bone marrow of these mice. Kinetics of engraftment differed between the two strains, in particular a significantly faster expansion of the human CD45(+) compartment and higher engraftment levels of CD3(+) T-cells were observed in NSG mice, which may explain the faster rate of GvHD development in this model. The pathogenesis of human GvHD involves anti-host effector cell reactivity and cutaneous tissue infiltration. Despite this, the presence of T-cell subsets and tissue homing markers has only recently been characterised in the peripheral blood of patients and has never been properly defined in Hu-PBMC models of GvHD. Engrafted human cells in NSG mice shows a prevalence of tissue homing cells with a T-effector memory (T(EM)) phenotype and high levels of cutaneous lymphocyte antigen (CLA) expression. Characterization of Hu-PBMC mice provides a strong preclinical platform for the application of novel immunotherapies targeting T(EM)-cell driven GvHD.

Choroidal neovascularisation (CNV) is a common vision-threatening complication of myopia and pathological myopia. Despite significant advances in understanding the epidemiology, pathogenesis and natural history of myopic CNV, there is no standard definition of myopic CNV and its relationship to axial length and other myopic degenerative changes. Several treatments are available to ophthalmologists, but with the advent of new therapies there is a need for further consensus and clinical management recommendations. Verteporfin photodynamic therapy has been an established treatment for subfoveal myopic CNV for many years, but this treatment does not restore visual acuity and is associated with long-term chorioretinal atrophy. More recently, clinical trials investigating the efficacy and safety of anti-vascular endothelial growth factor agents in patients with myopic CNV have demonstrated substantial visual acuity gains and quality of life increases compared with photodynamic therapy. These enhanced outcomes provide updated evidence-based clinical management guidelines of myopic CNV, and increase the need for a generally accepted definition for myopic CNV. This review critically summarises the latest myopic CNV literature in the context of clinical experience and recommends a myopic CNV treatment algorithm.

Remarkable advances in biomedical research have led to the generation of large amounts of data. Using artificial intelligence, it has become possible to extract meaningful information from large volumes of data, in a shorter frame of time, with very less human interference. In effect, convolutional neural networks (a deep learning method) have been taught to recognize pathological lesions from images. Diabetes has high morbidity, with millions of people who need to be screened for diabetic retinopathy (DR). Deep neural networks offer a great advantage of screening for DR from retinal images, in improved identification of DR lesions and risk factors for diseases, with high accuracy and reliability. This review aims to compare the current evidences on various deep learning models for diagnosis of diabetic retinopathy (DR).

An estimated 19 million of the world's children are visually impaired, while 1.4 million are blind. Using the UK as a model for high income countries, from a population-based incidence study, the annual cumulative incidence of severe visual impairment/blindness (SVL/BL) is estimated to be 6/10 000 by age 15 years, with the incidence being highest in the first year of life. The population of visually impaired children within high, middle and lower income countries differ considerably between and within countries. The numerous and mainly uncommon disorders which can cause impaired vision result in heterogeneous population which includes a substantial proportion (for SVI/BL, the majority) of children with additional systemic disorders or impairments whose needs differ substantially from those with isolated vision impairment. Paediatricians and other paediatric professionals have a key role in early detection and multidisciplinary management to minimise the impact of visual impairment (VI) in childhood.

BACKGROUND: Diabetic retinopathy screening in England involves labour-intensive manual grading of retinal images. Automated retinal image analysis systems (ARIASs) may offer an alternative to manual grading. OBJECTIVES: To determine the screening performance and cost-effectiveness of ARIASs to replace level 1 human graders or pre-screen with ARIASs in the NHS diabetic eye screening programme (DESP). To examine technical issues associated with implementation. DESIGN: Observational retrospective measurement comparison study with a real-time evaluation of technical issues and a decision-analytic model to evaluate cost-effectiveness. SETTING: A NHS DESP. PARTICIPANTS: Consecutive diabetic patients who attended a routine annual NHS DESP visit. INTERVENTIONS: Retinal images were manually graded and processed by three ARIASs: iGradingM (version 1.1; originally Medalytix Group Ltd, Manchester, UK, but purchased by Digital Healthcare, Cambridge, UK, at the initiation of the study, purchased in turn by EMIS Health, Leeds, UK, after conclusion of the study), Retmarker (version 0.8.2, Retmarker Ltd, Coimbra, Portugal) and EyeArt (Eyenuk Inc., Woodland Hills, CA, USA). The final manual grade was used as the reference standard. Arbitration on a subset of discrepancies between manual grading and the use of an ARIAS by a reading centre masked to all grading was used to create a reference standard manual grade modified by arbitration. MAIN OUTCOME MEASURES: Screening performance (sensitivity, specificity, false-positive rate and likelihood ratios) and diagnostic accuracy [95% confidence intervals (CIs)] of ARIASs. A secondary analysis explored the influence of camera type and patients' ethnicity, age and sex on screening performance. Economic analysis estimated the cost per appropriate screening outcome identified. RESULTS: A total of 20,258 patients with 102,856 images were entered into the study. The sensitivity point estimates of the ARIASs were as follows: EyeArt 94.7% (95% CI 94.2% to 95.2%) for any retinopathy, 93.8% (95% CI 92.9% to 94.6%) for referable retinopathy and 99.6% (95% CI 97.0% to 99.9%) for proliferative retinopathy; and Retmarker 73.0% (95% CI 72.0% to 74.0%) for any retinopathy, 85.0% (95% CI 83.6% to 86.2%) for referable retinopathy and 97.9% (95% CI 94.9 to 99.1%) for proliferative retinopathy. iGradingM classified all images as either 'disease' or 'ungradable', limiting further iGradingM analysis. The sensitivity and false-positive rates for EyeArt were not affected by ethnicity, sex or camera type but sensitivity declined marginally with increasing patient age. The screening performance of Retmarker appeared to vary with patient's age, ethnicity and camera type. Both EyeArt and Retmarker were cost saving relative to manual grading either as a replacement for level 1 human grading or used prior to level 1 human grading, although the latter was less cost-effective. A threshold analysis testing the highest ARIAS cost per patient before which ARIASs became more expensive per appropriate outcome than human grading, when used to replace level 1 grader, was Retmarker £3.82 and EyeArt £2.71 per patient. LIMITATIONS: The non-randomised study design limited the health economic analysis but the same retinal images were processed by all ARIASs in this measurement comparison study. CONCLUSIONS: Retmarker and EyeArt achieved acceptable sensitivity for referable retinopathy and false-positive rates (compared with human graders as reference standard) and appear to be cost-effective alternatives to a purely manual grading approach. Future work is required to develop technical specifications to optimise deployment and address potential governance issues. FUNDING: The National Institute for Health Research (NIHR) Health Technology Assessment programme, a Fight for Sight Grant (Hirsch grant award) and the Department of Health's NIHR Biomedical Research Centre for Ophthalmology at Moorfields Eye Hospital and the University College London Institute of Ophthalmology.

Compared to unaffected observers patients with schizophrenia (SZ) show characteristic differences in visual perception, including a reduced susceptibility to the influence of context on judgments of contrast - a manifestation of weaker surround suppression (SS). To examine the generality of this phenomenon we measured the ability of 24 individuals with SZ to judge the luminance, contrast, orientation, and size of targets embedded in contextual surrounds that would typically influence the target's appearance. Individuals with SZ demonstrated weaker SS compared to matched controls for stimuli defined by contrast or size, but not for those defined by luminance or orientation. As perceived luminance is thought to be regulated at the earliest stages of visual processing our findings are consistent with a suppression deficit that is predominantly cortical in origin. In addition, we propose that preserved orientation SS in SZ may reflect the sparing of broadly tuned mechanisms of suppression. We attempt to reconcile these data with findings from previous studies.

This paper considers how to identify statistical outliers in psychophysical datasets where the underlying sampling distributions are unknown. Eight methods are described, and each is evaluated using Monte Carlo simulations of a typical psychophysical experiment. The best method is shown to be one based on a measure of spread known as Sn. This is shown to be more sensitive than popular heuristics based on standard deviations from the mean, and more robust than non-parametric methods based on percentiles or interquartile range. Matlab code for computing Sn is included.

BACKGROUND/AIMS: To investigate outcomes following cataract surgery with and without primary intraocular lens (IoL) implantation in children under 2 years of age with congenital or infantile cataract. METHOD: Prospective population based cohort study undertaken through the British Isles Congenital Cataract Interest Group, with systematic data collection on children undergoing surgery in UK and Ireland between January 2009 and December 2010. ORs for the association between IoL implantation and visual acuity, postoperative glaucoma and reoperation at 1 year after surgery were estimated using multivariable regression analysis to control for potential confounders. RESULTS: Of 221 children, 56/131 with bilateral and 48/90 with unilateral cataract underwent primary IoL implantation. IoL implantation was independently associated with better visual outcome in bilateral (OR 4.6, 95% CI 1.6 to 13.1, p=0.004) but not unilateral disease. IoL use increased the odds of reoperation requiring repeat general anaesthetic (bilateral OR 5.5, p<0.01; unilateral OR 16.7, p<0.01). IoL implantation did not reduce the odds of postoperative glaucoma. CONCLUSIONS: The use of IoLs in cataract surgery in young children should be critically reassessed, particularly used in settings/communities where close, long-term follow-up is challenging. The absence of visual benefit and the lack of a previously postulated protective effect against postoperative glaucoma serve to question the value of IoLs in unilateral disease. The potential association between IoL use and better early visual outcomes in bilateral disease needs to be balanced against the risk of reoperation and exposure to additional general anaesthetics during a key period of neurodevelopment.

IMPORTANCE: The comparative clinical effectiveness of ranibizumab, aflibercept, and bevacizumab for the management of macular edema due to central retinal vein occlusion (CRVO) is unclear. OBJECTIVE: To determine whether intravitreal aflibercept or bevacizumab compared with ranibizumab results in a noninferior mean change in vision at 100 weeks for eyes with CRVO-related macular edema. DESIGN, SETTING, AND PARTICIPANTS: This prospective, 3-arm, double-masked, randomized noninferiority trial (Lucentis, Eylea, Avastin in Vein Occlusion [LEAVO] Study) took place from December 12, 2014, through December 16, 2016, at 44 UK National Health Service ophthalmology departments. Inclusion criteria included age 18 years or older, visual impairment due to CRVO-related macular edema of less than 12 months with best-corrected visual acuity (BCVA) Early Treatment Diabetic Retinopathy Study letter score (approximate Snellen equivalent) in the study eye between 19 (20/400) and 78 (20/32), and spectral domain optical coherence tomography imaging central subfield thickness of 320 μm or greater. Data were analyzed from March 4, 2019, to April 26, 2019. INTERVENTIONS: Participants were randomized (1:1:1) to receive repeated intravitreal injections of ranibizumab (0.5 mg/0.05 mL) (n = 155), aflibercept (2.0 mg/0.05 mL) (n = 154), or bevacizumab (1.25 mg/0.05 mL) (n = 154) for 100 weeks. MAIN OUTCOMES AND MEASURES: Adjusted mean change in BCVA in the study eye at 100 weeks wherein noninferiority was concluded if the lower bounds of the 95% CI of both the intention-to-treat and the per protocol analyses were above -5 letters. RESULTS: Of 463 participants, 265 (57.2%) were male, with a mean (SD) age of 69.1 (13.0) years. The mean (SD) gain in BCVA letter score was 12.5 (21.1) for ranibizumab, 15.1 (18.7) for aflibercept, and 9.8 (21.4) for bevacizumab at 100 weeks. Aflibercept was noninferior to ranibizumab (intention-to-treat-adjusted mean BCVA difference, 2.23 letters; 95% CI, -2.17 to 6.63 letters; P < .001). Bevacizumab was not noninferior to ranibizumab (intention-to-treat-adjusted mean BCVA difference, -1.73 letters; 95% CI, -6.12 to 2.67 letters; P = .07). The per protocol analysis conclusions were similar. Fewer mean injections were given in the aflibercept group (10.0) than in the ranibizumab (11.8) group (mean difference at 100 weeks, -1.9; 95% CI, -2.9 to -0.8). CONCLUSIONS AND RELEVANCE: Mean changes in vision after treatment of macular edema due to CRVO were no worse using aflibercept compared with ranibizumab. Mean changes in vision using bevacizumab compared with ranibizumab were inconclusive regarding vision outcomes (ie, the change in visual acuity from baseline, on average, may be worse or may not be worse when using bevacizumab compared with ranibizumab). TRIAL REGISTRATION: ISRCTN13623634.

BACKGROUND: Disorganization of the retinal inner layers (DRIL) on optical coherence tomography (OCT) is thought to represent retinal capillary non-perfusion (CNP) in eyes with diabetic retinopathy. This study was designed to evaluate the ability of DRIL to accurately predict CNP. DESIGN: Retrospective masked reliability and diagnostic accuracy study performed in the National Institute for Health Research (NIHR) Moorfields Biomedical Research Centre, London, UK SAMPLES: Retinal images of patients with diabetic retinopathy METHODS: The OCT images from 90 separate areas of angiographically confirmed perfused and non-perfused areas of the macula from 37 eyes of 31 patients were anonymized and coded. Two masked graders independently graded these OCT scans for the presence or absence of DRIL to determine the intergrader reliability. The diagnostic accuracy of DRIL in identifying CNP was evaluated from the results obtained. MAIN OUTCOME MEASURES: Sensitivity and specificity of DRIL in accurately detecting CNP RESULTS: The intergrader agreement was high with a Cohen's kappa of 0.909. DRIL was present in 84.4% (38/45) of non-perfused retina and none in perfused retina (0/45). The sensitivity and specificity of DRIL in detecting angiographic evidence of CNP was 84.4% and 100%, respectively. The positive predictive value was 100% and the negative predictive value was 86.5%. CONCLUSIONS: The presence of DRIL is a reliable predictor of areas of macular CNP. However, DRIL is not a universal finding of non-perfusion, with some cases exhibiting absence of DRIL despite angiographic CNP.

Purpose: To develop and assess a method for predicting the likelihood of converting from early/intermediate to advanced wet age-related macular degeneration (AMD) using optical coherence tomography (OCT) imaging and methods of deep learning. Methods: Seventy-one eyes of 71 patients with confirmed early/intermediate AMD with contralateral wet AMD were imaged with OCT three times over 2 years (baseline, year 1, year 2). These eyes were divided into two groups: eyes that had not converted to wet AMD (n = 40) at year 2 and those that had (n = 31). Two deep convolutional neural networks (CNN) were evaluated using 5-fold cross validation on the OCT data at baseline to attempt to predict which eyes would convert to advanced AMD at year 2: (1) VGG16, a popular CNN for image recognition was fine-tuned, and (2) a novel, simplified CNN architecture was trained from scratch. Preprocessing was added in the form of a segmentation-based normalization to reduce variance in the data and improve performance. Results: Our new architecture, AMDnet, with preprocessing, achieved an area under the receiver operating characteristic (ROC) curve (AUC) of 0.89 at the B-scan level and 0.91 for volumes. Results for VGG16, an established CNN architecture, with preprocessing were 0.82 for B-scans/0.87 for volumes versus 0.66 for B-scans/0.69 for volumes without preprocessing. Conclusions: A CNN with layer segmentation-based preprocessing shows strong predictive power for the progression of early/intermediate AMD to advanced AMD. Use of the preprocessing was shown to improve performance regardless of the network architecture.

PURPOSE: To evaluate the potential role of hyperreflective foci (HF) in predicting visual outcome in patients undergoing treatment for macular edema due to retinal vascular diseases. METHODS: Data and images of 92 patients with macular edema due to diabetes mellitus or branch retinal vein occlusion, treated with either intravitreal dexamethasone implant or ranibizumab, were analyzed. All patients underwent best-corrected visual acuity measurement, slit-lamp examination, spectral domain optical coherence tomography at baseline and at all time points of the follow-up (Month 1, 2, 3, 6, and 9). Generalized least squares random effects linear or logistic regression analysis was used to investigate potential factors associated with the final best-corrected visual acuity and number of HF, respectively. RESULTS: Increasing age (P < 0.001), central retinal thickness (P < 0.001), number of HF (P = 0.028), presence of subretinal fluid (P < 0.001), intraretinal fluid (P < 0.001), intraretinal cysts (P < 0.001), and disruption of ellipsoid zone/external limiting membrane (P < 0.001) were significantly associated with poorer visual outcome. Factors associated with HF were increasing central retinal thickness (P = 0.003), presence of subretinal fluid (P = 0.049), intraretinal fluid (P = 0.002), cysts (P = 0.015), and disruption of ellipsoid zone (P = 0.047). No significant differences in change in best-corrected visual acuity, central retinal thickness, and HF were observed between the two treatment groups. CONCLUSION: Hyperreflective foci are associated with poorer visual outcome in patients with macular edema due to retinal vascular diseases. Similar reductions in HF are achieved by intravitreal steroid and anti-vascular endothelial growth factor agent.