AIMS: To investigate the changes in choroidal optical coherence tomography (OCT) radiomic features, their correlations with visual acuity and utility in identifying pathological myopia (PM). METHODS: A total of 288 myopi...AIMS: To investigate the changes in choroidal optical coherence tomography (OCT) radiomic features, their correlations with visual acuity and utility in identifying pathological myopia (PM). METHODS: A total of 288 myopic participants aged 18-50 years were included. Choroidal radiomic features were extracted and screened from OCT images via PyRadiomics and machine learning. Selected features were analysed for their associations with axial length, spherical equivalent, age and best corrected visual acuity (BCVA). Classification models were built using these features and their performance to identify PM was evaluated and compared with clinical parameters through five-fold cross-validation using metrics including area under the curve (AUC), accuracy, recall and F1 score. RESULTS: A total of 464 radiomic features were extracted and four choroidal intensity and shape features significantly correlated with axial length (p<0.001) were selected. Smaller maximum diameter, higher coarseness and greater perimeter surface ratio were significantly associated with worse BCVA (p<0.05). Radiomic features showed better performance than clinical features in both the internal test set (AUC=0.970 vs 0.938) and external validation set (AUC=0.858 vs 0.753) in identifying PM. Combining radiomic features with risk factors improved classification performance (AUC=0.990 internally and 0.892 externally). Among the included factors, intensity feature was most predictive for PM. CONCLUSIONS: OCT-based choroidal intensity and shape features were significantly correlated with axial elongation and visual impairment and outperformed clinical parameters in identifying PM. These features could serve as reliable biomarkers for monitoring high myopia progression.
Romano F, Stettler I, Finn M
… +17 more, Vingopoulos F, Ding X, Overbey K, Cort T, Chen C, Ploumi I, Baldwin G, Zhu Y, Nodecker KN, Gumustop SS, Shah SH, Laíns I, Kim L, Vavvas D, Husain D, Miller JW, Miller JB
AIMS: To investigate associations between imaging biomarkers and quantitative contrast sensitivity (CS) function (qCSF) metrics in geographic atrophy (GA). METHODS: Cross-sectional study including 97 eyes from 70 patient...AIMS: To investigate associations between imaging biomarkers and quantitative contrast sensitivity (CS) function (qCSF) metrics in geographic atrophy (GA). METHODS: Cross-sectional study including 97 eyes from 70 patients (>55 years) with GA within 1500 µm of the fovea and visual acuity (VA)>20/320. Participants underwent VA and qCSF testing (area under the log CS function (AULCSF), contrast acuity (CA), and CS at 1-18 cycles per degree (cpd)) and multimodal imaging with blue autofluorescence (BAF), optical coherence tomography (OCT), and, in 55 eyes, swept-source OCT angiography (SS-OCTA). Imaging biomarkers included GA size, prior growth rate, configuration, BAF pattern, circularity, percentage of foveal involvement (%FI), foveal distance and macular inner choroid flow deficit percentage (mICFD%). Generalised linear mixed-effects and random forest (GRF) models evaluated structure-function relationships. RESULTS: Participants (age 78.8±5.6 years; 70% female) had a median GA size of 4.0 mm², with 64% showing subfoveal involvement. Larger GA size and higher %FI were significantly associated with worse VA, AULCSF, CA and CS at 1-12 cpd (all p<0.01). Unifocal lesions correlated with lower VA and CS at 6 cpd, while diffuse/banded BAF patterns and greater mICFD% were linked to reduced CS at 1-1.5 cpd. GRF analysis identified GA size and %FI as primary predictors of CS loss, with best performance for AULCSF and CS at 3 cpd (R²=0.319, 0.314). CONCLUSIONS: GA size and %FI are key determinants of CS loss in GA. qCSF outperformed VA in predictive accuracy, supporting its use as a sensitive functional endpoint in GA trials.
BACKGROUND/AIMS: To assess the intergrader variability in detecting incomplete outer retinal atrophy (iRORA), an optical coherence tomography (OCT) biomarker for early atrophic changes in age-related macular degeneration...BACKGROUND/AIMS: To assess the intergrader variability in detecting incomplete outer retinal atrophy (iRORA), an optical coherence tomography (OCT) biomarker for early atrophic changes in age-related macular degeneration (AMD) and a potential clinical trial endpoint. METHODS: We performed a cross-sectional intergrader agreement study using 60 OCT B-scans enriched for iRORA features from non-exudative AMD eyes. Five international retinal experts independently graded images for iRORA presence according to the current Classification of Atrophy Meeting (CAM) criteria, using a standardised online annotation platform. Pairwise agreement was assessed with Cohen's Kappa; agreement with the majority vote, sensitivity and specificity were calculated using a leave-one-out approach. Annotation lead times were recorded to explore links between decision speed and grading consistency. RESULTS: Mean pairwise Cohen's Kappa was 0.425 (range 0.19-0.73), indicating mild agreement. Agreement with the majority vote ranged from κ=0.48 to 0.67, with classification accuracy between 0.73 and 0.83. Sensitivity varied from 0.59 to 0.83, while specificity remained high (0.85-0.96). Annotation lead times were generally shorter when grader decisions aligned with the majority consensus. CONCLUSION: Despite standardised CAM definitions, intergrader reproducibility for iRORA detection remains limited, reflecting challenges in consistent early atrophy identification. Refinement of grading guidelines and integration of automated detection systems could improve reliability for AMD clinical trials and clinical practice.
Prophylactic treatment of lattice degeneration with laser photocoagulation remains controversial. We evaluated whether prophylactic laser for lattice degeneration increases epiretinal membrane (ERM) risk. In a retrospect...Prophylactic treatment of lattice degeneration with laser photocoagulation remains controversial. We evaluated whether prophylactic laser for lattice degeneration increases epiretinal membrane (ERM) risk. In a retrospective cohort (2005-2025), patients receiving laser within 90 days of lattice diagnosis were classified as exposed; untreated patients served as controls and were matched 3:1. ERM within 10 years was analysed using Fine-Gray competing-risk models (death as competing event). ERM occurred in 38/321 (11.8%) laser-treated versus 53/944 (5.6%) untreated patients (sHR=2.50, 95% CI 1.65 to 3.79; p<0.001). Prophylactic laser was associated with higher ERM risk, warranting careful consideration of its risk-benefit profile.
BACKGROUND/AIMS: To identify predictors of continued discontinuation of vascular endothelial growth factor (VEGF) inhibitors for retinal vein occlusion (RVO) after it has been suspended. METHODS: Eyes with RVO were ident...BACKGROUND/AIMS: To identify predictors of continued discontinuation of vascular endothelial growth factor (VEGF) inhibitors for retinal vein occlusion (RVO) after it has been suspended. METHODS: Eyes with RVO were identified in the Fight Retinal Blindness! registry that suspended VEGF inhibitors for ≥180 days when visual acuity (VA) was ≥35 letters, from 1 December 2011-2021. The primary outcome was not resuming therapy through 24 months which defined success. RESULTS: Half (186/362 (51%)) of the eligible 6-month suspensions successfully continued through 24 months. Successful cessation attempts began earlier (9 vs 14 months; p=0.001), after greater VA improvement (mean, +12.6 vs +8.5 letters; p=0.01) following more injections in the first 6 months (median, 5 vs 4; p<0.001) and more intensive treatment than unsuccessful attempts (176/362 (49%)). At month 6 without treatment, the mean (95% CI) changes in VA and central subfield thickness (CST) were in successful attempts +1.8 letters (+0.4 to +3.1) and -20 µm (-34 to -6); unsuccessful attempts +0.5 letters (-1.1 to 2.0) and -2 µm (-22 to +18). When therapy resumed in unsuccessful attempts (median, 9.5 months), VA and CST had deteriorated by -3.5 letters (-5.6 to -1.4) and +38 µm (+20 to +57) since suspension. These trends were more pronounced in eyes with central retinal vein occlusion. CONCLUSION: Eyes with RVO that achieve good outcomes early, following more intensive VEGF inhibition and that improve further over 6 months without therapy are likely to achieve sustained, treatment-free quiescence through 24 months.
Lu F, Jiang H, Wong W
… +16 more, Ren N, Hu J, Zhao J, Jin L, Zhang E, Zhang G, Lian J, Chen L, Chen M, He C, Fu J, Chen Q, Chen X, Ding L, Han X, Liang L
BACKGROUND: Uncorrected refractive error (URE) is the leading cause of vision impairment (VI). Effective refractive error coverage (eREC) is a WHO-endorsed indicator for tracking the global eyecare progress; however, the...BACKGROUND: Uncorrected refractive error (URE) is the leading cause of vision impairment (VI). Effective refractive error coverage (eREC) is a WHO-endorsed indicator for tracking the global eyecare progress; however, there is a lack of reliable eREC estimate for children due to a paucity of available data. This study aimed to estimate the eREC and prevalence of VI among schoolchildren in Xinjiang, China. METHODS: A population-based, cross-sectional study was conducted among schoolchildren aged 7-20 years from May to June 2024. Uniocular uncorrected, presenting and best-corrected visual acuity was measured using standard logarithmic vision charts per standardised protocol, and refractive error coverage indicators (eREC, refractive error coverage (REC) and Quality Gap) were calculated according to WHO definitions. Sociodemographic correlates of eREC and REC were assessed using multivariable logistic regression analysis. RESULTS: A total of 3306 children were examined (mean age, 12.5±2.3 years; 50.03% boys). The prevalence of VI based on uncorrected, presenting and best-corrected visual acuity was 8.74%, 4.30% and 0.15%, respectively. The overall eREC and REC were 51.76% and 54.93%, with a quality gap of 5.77%. Age and sex did not show a significant association with eREC, whereas secondary school education (vs primary school education, OR=3.59, 95% CI 1.52 to 9.29, p=0.005) was significantly associated with a higher eREC. CONCLUSION: 96.51% of VI were due to URE among school-aged children in Xinjiang, China. The eREC remains well below the WHO 2030 global target for China, highlighting the need to improve both the accessibility and quality of refractive correction in this population.
BACKGROUND: Postoperative microbial keratitis (MK) can cause graft failure and vision loss following keratoplasty for corneal endothelial decompensation. Many previous studies found penetrating keratoplasty (PK) to resul...BACKGROUND: Postoperative microbial keratitis (MK) can cause graft failure and vision loss following keratoplasty for corneal endothelial decompensation. Many previous studies found penetrating keratoplasty (PK) to result in a higher incidence of MK than Descemet's membrane endothelial keratoplasty (DMEK) or Descemet's stripping automated endothelial keratoplasty (DSAEK); however, findings vary. Recently, there has been a shift towards selective transplant procedures, with DMEK becoming mainstream over PK for endothelial decompensation globally around 2017. This systematic review therefore aims to determine the incidence of MK following DMEK, DSAEK and PK for endothelial decompensation from 2017 to 2025 and to assess variation by follow-up duration, socioeconomic status, microbe type, patient age, contact lens wear and time after surgery. METHODS: Multiple databases were systematically searched during 2017-2025. The Newcastle-Ottawa scale evaluated bias. Random-effects meta-analyses with logit transformation were performed; heterogeneity was assessed using the I statistic. Sensitivity analyses confirmed robustness. RESULTS: Pooled MK incidence was highest following PK (4.84%; 95% CI 3.25% to 7.15%; I=97.7%), followed by DMEK (0.71%; 95% CI 0.18% to 2.86%; I=83.5%), then DSAEK (0.27%; 95% CI 0.19% to 0.38%; I=0%). Subgroup analyses showed variation but few significant differences. Trends suggested significantly higher MK incidence following PK in studies with longer follow-up. Bacterial keratitis was the most common overall; however, DMEK showed a more even distribution across microbe types. CONCLUSION: MK incidence was highest following PK. High heterogeneity and limited DMEK/DSAEK data warrant cautious interpretation. PROSPERO REGISTRATION NUMBER: CRD420251179122.
BACKGROUND/AIMS: To evaluate associations of sociodemographic and functional parameters with vision-related quality of life (VRQoL) in recessive Stargardt disease (STGD1). METHODS: A total of 71 participants (42 females,...BACKGROUND/AIMS: To evaluate associations of sociodemographic and functional parameters with vision-related quality of life (VRQoL) in recessive Stargardt disease (STGD1). METHODS: A total of 71 participants (42 females, 29 males; mean age 44±19 years) with genetically confirmed STGD1 were included in this cohort study. Two validated patient-reported outcome measures (PROMs), namely National Eye Institute Visual Function Questionnaire and Impact of Vision Impairment profile, were administered to the participants. Responses were analysed using latent trait models following psychometrically established dimension structures (functional and emotional subscales). Univariable and linear mixed-effects models were applied to investigate the association of putative determinants with VRQoL. RESULTS: The optimised models could predict the measured VRQoL impairment up to a multicollinearity-corrected adjusted accuracy of 0.558. Functional subscales could more accurately be predicted than emotional subscales. Overall, reading acuity was the most important determinant of VRQoL. Other functional parameters, including visual function of the worse eye, revealed significant impact as well while the influence of sociodemographic parameters on VRQoL was more inconsistent. CONCLUSIONS: The robust associations between VRQoL and visual function in STGD1 indicate that both PROM are suitable and construct valid outcome measures for upcoming interventional trials. Future clinical trials and patient assessment focusing on VRQoL might take near vision of both eyes into consideration.
AIMS: To evaluate whether predominantly peripheral lesions (PPLs) and other imaging biomarkers on ultra-widefield (UWF) photography can predict the progression of diabetic retinopathy (DR) in a multiethnic Asian cohort....AIMS: To evaluate whether predominantly peripheral lesions (PPLs) and other imaging biomarkers on ultra-widefield (UWF) photography can predict the progression of diabetic retinopathy (DR) in a multiethnic Asian cohort. METHODS: This was a prospective, longitudinal cohort study involving 282 participants (528 eyes) with diabetes and either no DR or non-proliferative DR, recruited from the Singapore National Eye Centre between July 2017 and May 2021. Participants underwent annual UWF colour fundus photography and systemic evaluations over a 2-year period. Images were graded at a centralised reading centre. An automated lesion detection algorithm was used for objective quantification of microaneurysms and retinal haemorrhages. Multivariate regression analysis was conducted to identify independent predictors of DR progression, adjusting for systemic and ocular risk factors. The primary outcome was DR progression, defined as a ≥2 step worsening on the Diabetic Retinopathy Severity Scale or development of proliferative DR within 2 years. RESULTS: The 2-year progression rate of DR was 9.85%. Independent predictors of progression included increased peripheral retinal haemorrhage density (OR=3.09; 95% CI 1.03 to 9.22; p=0.044), presence of PPLs (OR=4.00; 95% CI 1.07 to 15.0; p=0.040) and higher baseline diastolic blood pressure (OR=1.04; 95% CI 1.01 to 1.08; p=0.015). Eyes with PPLs had a 1.6-fold higher risk of progression compared with eyes with predominantly central lesions (15.3% vs 9.3%). CONCLUSION: Peripheral biomarkers on UWF imaging, including peripheral retinal haemorrhage density and PPLs, are independent predictors of DR progression. These findings support the clinical utility of peripheral retinal assessment and automated artificial intelligence-based imaging tools in DR risk stratification and management.
BACKGROUND/AIMS: Secondary intraocular lens (IOL) implantation is reported to increase glaucoma development and progression in paediatric patients. We evaluated the risk of new-onset glaucoma or progression following sec...BACKGROUND/AIMS: Secondary intraocular lens (IOL) implantation is reported to increase glaucoma development and progression in paediatric patients. We evaluated the risk of new-onset glaucoma or progression following secondary IOL implantation in children and young adults. METHODS: Retrospective cohort study of secondary IOL implantation following paediatric cataract surgery at Moorfields Eye Hospital (2017-2025). Anterior chamber IOLs or retinal pathology were excluded. Primary outcome was glaucoma progression (worsening optic disc appearance, sustained intraocular pressure (IOP) elevation requiring medical therapy escalation and/or glaucoma surgery). Secondary outcomes included longitudinal IOP trends and best-corrected visual acuity. RESULTS: 57 eyes were included (30 paediatric, 0-16 years; 27 young adult, 17-30 years); mean follow-up 28.2 months (SD±26.4). Aetiology was congenital (49%), uveitic (26%) and traumatic (25%). IOL position was ciliary sulcus (68%), capsular bag (18%) and scleral-fixated (14%). Glaucoma progression occurred in 16 eyes (28%), predominantly in eyes with pre-existing glaucoma (15/21, 71%), vs 1/36 eyes (3%) without pre-existing glaucoma (relative risk 25.7; 95% CI 3.7 to 181.0; p<0.0001). Progression was most frequent in uveitic eyes (67%) and was not observed in traumatic cataract (p=0.00012). No significant association was observed between age group or IOL position and glaucoma progression. CONCLUSIONS: Secondary IOL implantation is associated with a significant risk of glaucoma progression in eyes with pre-existing glaucoma but appears unlikely to induce new-onset glaucoma in eyes without prior disease. These findings highlight the importance of careful patient selection, counselling and close postoperative surveillance in patients with pre-existing glaucoma undergoing secondary IOL implantation.
BACKGROUND/AIMS: PINNACLE is one of the largest prospective multicentre observational studies evaluating the progression of intermediate age-related macular degeneration (iAMD). This paper aims to provide an optical cohe...BACKGROUND/AIMS: PINNACLE is one of the largest prospective multicentre observational studies evaluating the progression of intermediate age-related macular degeneration (iAMD). This paper aims to provide an optical coherence tomography (OCT)-based qualitative and quantitative characterisation of the cohort's baseline morphology. METHODS: Based on expert grader readings and artificial intelligence (AI)-based image analysis, we report the prevalence, quantitative measurements, topographic distribution and intercorrelation of characteristic iAMD features including drusen, drusen subtypes, subretinal drusenoid deposits (SDD), hyperreflective foci (HRF), double layer sign and various measurements of outer retinal condition, such as ellipsoid zone (EZ) and outer nuclear layer (ONL) thicknesses, incomplete retinal pigment epithelium and outer retinal atrophy (iRORA) and a semi-automated atrophic marker including EZ loss and choroidal hypertransmission (EZLHT). RESULTS: Drusen accumulate within the central 3 mm while SDD predominantly occurs in the superior perifoveal quadrant. Whereas higher drusen volume was associated with the presence and volume of HRF, it was inversely correlated with SDD presence. Thickness measurements of the EZ and ONL demonstrate outer retinal thinning, indicating photoreceptor compromise in iAMD, more pronounced in eyes showing atrophic features such as iRORA or EZLHT. CONCLUSION: This work combines expert grader readings with AI-based image analyses, applied on the largest prospectively, densely OCT-imaged cohort of iAMD reported on so far. The results show feature distribution comparable to previous reports. They substantially contribute to the comprehensive morphological characterisation of iAMD. This data is relevant for the interpretation of longitudinal data, refining inclusion criteria for future clinical trials and for providing a reference to other trials in the field.
PURPOSE: The cross-sectional and longitudinal associations between accelerated biological ageing and the risk of cataract and other blinding eye diseases (including glaucoma and age-related macular degeneration (AMD)) re...PURPOSE: The cross-sectional and longitudinal associations between accelerated biological ageing and the risk of cataract and other blinding eye diseases (including glaucoma and age-related macular degeneration (AMD)) remain unclear. METHODS: We included participants aged 40 and above with biological ages, including phenotypic age (PhenoAge), Klemera-Doubal method (KDMAge) and retinal age (RetiAge), from the US National Health and Nutrition Examination Survey (NHANES) and UK Biobank. The cross-sectional analyses were conducted to identify associations of PhenoAge or KDMAge acceleration with cataract and other blinding eye diseases using logistic regression. In a prospective UK cohort, we explored the relationships between the acceleration of PhenoAge, KDMAge or RetiAge and cataract and other blinding eye diseases using the Cox proportional hazards model. RESULTS: This study consisted of 5433 participants from the US NHANES and 269 615 participants from the UK Biobank. In both cross-sectional and longitudinal analyses, accelerated biological ageing was positively associated with an increased risk of cataract (all p<0.05). In a longitudinal cohort, RetiAge acceleration demonstrated the larger effect size estimates (HR 1.54 (95% CI 1.38 to 1.73)) compared with PhenoAge acceleration (HR 1.05 (95% CI 1.03 to 1.08)) and KDMAge acceleration (HR 1.06 (95% CI 1.04 to 1.08)). Only in the UK population, risks of glaucoma showed stronger links with KDMAge acceleration (HR 1.06 (95% CI 1.01 to 1.11)), while AMD showed more pronounced associations with PhenoAge acceleration (HR 1.08 (95% CI 1.02 to 1.14)). CONCLUSIONS: Accelerated biological ageing might represent a potential target of assessment and intervention for cataract.
PURPOSE: To determine whether posterior staphyloma (PS) is associated with long-term trajectories of ocular structural changes, myopia-related complications and visual outcomes in high myopia. METHODS: This prospective l...PURPOSE: To determine whether posterior staphyloma (PS) is associated with long-term trajectories of ocular structural changes, myopia-related complications and visual outcomes in high myopia. METHODS: This prospective longitudinal cohort study included 614 highly myopic eyes from the Zhongshan High Myopia Cohort Study, followed for up to 12 years. Eyes were categorised by PS status: baseline PS (n=46), new-onset PS (n=94) or no PS (n=474). The associations between PS and myopic macular degeneration (MMD) progression, myopic traction maculopathy (MTM) and plus lesions incidence, longitudinal changes in ocular parameters and incident moderate-to-severe visual impairment (MSVI) were assessed. RESULTS: PS was independently associated with increased risks of MMD progression (OR 3.39, 95% CI 1.81 to 6.35, p<0.001), MTM (OR 2.25, 95% CI 1.06 to 4.77, p=0.034) and incident plus lesions (OR 3.15, 95% CI 1.16 to 8.57, p=0.024). The highest risks of MTM and plus lesions were observed in eyes with new-onset PS. Eyes with PS demonstrated significantly faster axial elongation (mean annual rate: baseline PS, 0.120 mm (95% CI 0.099 to 0.142); new-onset PS: 0.129 mm (95% CI 0.114 to 0.145); no PS: 0.066 mm (95% CI 0.060 to 0.073), p<0.001) and a fivefold higher risk of MSVI (95% CI 1.36 to 18.45, p=0.016). CONCLUSIONS: PS marks a pivotal structural transition in high myopia, identifying a phase of accelerated axial elongation, increased risk of complications and worse visual outcomes, and may serve as a key marker for risk stratification and an important target for early intervention.
AIMS: To evaluate associations between systemic inflammation biomarkers and incident age-related ocular diseases while also investigating their correlations with retinal structures. METHODS: This population-based prospec...AIMS: To evaluate associations between systemic inflammation biomarkers and incident age-related ocular diseases while also investigating their correlations with retinal structures. METHODS: This population-based prospective cohort study analysed 415 599 UK Biobank participants. Systemic immune-inflammation index (SII) and low-grade inflammation score (INFLA-score) were calculated from baseline haematological parameters. Primary outcomes were incident diagnoses of cataract, primary open-angle glaucoma (POAG), age-related macular degeneration (AMD) and diabetic retinopathy (DR). Multivariable Cox proportional hazards models estimated HRs with 95% CIs. Secondary analyses assessed the associations with optical coherence tomography-derived retinal layer thicknesses and vascular features. RESULTS: Over a median 13.0-year follow-up, we identified 44 906 cataract, 5803 POAG, 7388 AMD and 3319 DR incident cases. Both SII and INFLA-score demonstrated significant, dose-dependent associations with all ocular outcomes (all p<0.05). Distinct exposure-response patterns emerged: J-shaped relationships for cataract and POAG (SII threshold >500; INFLA-score threshold >0), versus monotonically positive associations for AMD and DR. Elevated inflammatory markers also correlated with retinal thinning, especially in photoreceptor layers. CONCLUSIONS: Systemic inflammation biomarkers could predict incident age-related ocular diseases with disease-specific patterns while concurrently associating with quantifiable retinal structural and vascular pathologies. These findings suggest that anti-inflammatory strategies might have potential to mitigate ocular ageing processes, although further evidence on causal mechanisms and interventions is warranted.
Fraser-Bell S, Lee NS, Hashimoto Y
… +10 more, Pollreisz A, Witmer A, Viola F, Lavid FJ, Steiner H, Gabrielle PH, Barry R, Silva R, Barthelmes D, Gillies MC
AIMS: To compare 1-year and 2-year outcomes of eyes with persistent and non-persistent diabetic macular oedema (DME) treated with vascular endothelial growth factor (VEGF) inhibitors. METHODS: This was a cohort study usi...AIMS: To compare 1-year and 2-year outcomes of eyes with persistent and non-persistent diabetic macular oedema (DME) treated with vascular endothelial growth factor (VEGF) inhibitors. METHODS: This was a cohort study using the Fight Retinal Blindness! (FRB!) international outcomes registry. Participants had treatment-naïve eyes with centre-involving DME starting intravitreal VEGF inhibitor therapy from 2014 to 2023. Eyes were grouped as those with persistent DME and those with non-persistent DME. Main outcome measures were mean visual acuity (VA) change and central subfoveal thickness change after 1 and 2 years of treatment. RESULTS: This study included 877 eyes, of which 40% had persistent DME. The mean VA change in eyes with persistent DME was less than that in eyes with non-persistent DME at 1 (+3.6 vs. +6.5 letters; p<0.001) and 2 (+3.6 vs. +6.3 letters; p<0.001) years. The persistent group had a lower mean reduction in central subfield thickness at 1 (-74 vs -111 µm; p<0.001) and 2 (-90 vs -114 µm; p<0.001) years and received more mean injections at 1 (7.6 vs 6.5; p<0.001) and 2 (11.8 vs 9.5; p<0.001) years. CONCLUSION: Eyes with persistent DME had significantly lower improvement in mean VA, lower reduction in CST and more injections 1 and 2 years after the initiation of VEGF inhibitor therapy. Eyes with persistent DME likely had a higher treatment burden because they had more aggressive disease that was more difficult to control. More effective agents would likely deliver better outcomes in this significant group of patients with DME.
AIMS: This study evaluated the use of ophthalmic foundation deep-learning models with cross-modal transfer learning to classify multiple diseases on optical coherence tomography angiography (OCTA) with limited sample siz...AIMS: This study evaluated the use of ophthalmic foundation deep-learning models with cross-modal transfer learning to classify multiple diseases on optical coherence tomography angiography (OCTA) with limited sample size. METHODS: The OCTA-500 dataset (n=500 subjects) was split into an 85% training/validation set for fivefold cross-validation and a 15% held-out test set. Superficial and deep projections from OCTA were combined using intermediate fusion. Outcomes were multi-disease classification of normal, diabetic retinopathy, age-related macular degeneration and 'other'. Transfer-learning from colour fundus photography was used to overcome the limitation of small sample sizes. Vision-Transformer-VisionFM and Vision-Transformer-RETFound domain-specific foundation models with cross-modal transfer learning were evaluated. Comparison was made with Vision-Transformer-ImageNet, a non-domain-specific model. Performance was evaluated using accuracy, F1-score, precision, recall and area under the receiver operating characteristic curve. Saliency maps were also computed. RESULTS: VisionFM with cross-modal transfer learning outperformed ImageNet in classifying different diseases on OCTA (accuracy: 0.8133±0.0470 vs 0.7600±0.0502). RETFound with cross-modal transfer learning outperformed ImageNet in classifying different diseases on OCTA (accuracy: 0.8000±0.0507 vs 0.7600±0.0521). Similar conclusions were reached with other performance metrics. Saliency maps from VisionFM and RETFound yielded attention patterns that localised pathology to relevant retinal structures on superficial and deep projections from OCTA, comparing favourably with those from ImageNet models. CONCLUSIONS: Retinal foundation models with cross-modal transfer learning enable accurate multi-class classification using OCTA data, which had small sample size. Results from domain-specific foundation models compared favourably with a non-domain-specific model. Saliency analysis showed attention patterns of pathology localised to anatomically relevant retinal structures.
AIM: This study aims to develop and validate machine learning models for predicting recurrence in polypoidal choroidal vasculopathy (PCV) patients using optical coherence tomography (OCT) and OCT angiography (OCTA) bioma...AIM: This study aims to develop and validate machine learning models for predicting recurrence in polypoidal choroidal vasculopathy (PCV) patients using optical coherence tomography (OCT) and OCT angiography (OCTA) biomarkers. METHODS: This multicentre prospective study was conducted at 14 hospitals between June 2019 and December 2023. Patients who achieved remission after anti-vascular endothelial growth factor treatment were followed up for at least 1 year with serial OCT/OCTA imaging. Predictive features were selected using the least absolute shrinkage and selection operator (LASSO) regression with 10-fold cross-validation. Five classifiers (logistic regression, support vector machine, random forest, k-nearest neighbours and extreme gradient boosting (XGBoost)) were developed and evaluated using area under the curve (AUC), accuracy, sensitivity and specificity. Shapley additive explanations (SHAP) were applied for model interpretation and feature ranking. RESULTS: A total of 204 eyes were included, with 125 from Peking Union Medical College Hospital (training set) and 79 from 13 other centres across China (external validation set). Ten features were selected for model development. In the external validation set, AUCs ranged from 0.801 to 0.861, with the XGBoost model achieving the highest AUC (0.861). SHAP analysis revealed that the percent change in polyp height, the change in branching neovascular network area and the change in subfoveal choroidal thickness were the top three significant predictors. CONCLUSION: The XGBoost model demonstrated the best predictive performance, providing a reliable tool for recurrence prediction, aiding personalised treatment decisions and optimising clinical resources.
BACKGROUND: Diabetic retinopathy (DR) is a leading cause of preventable blindness among people with type 2 diabetes mellitus (T2DM). Early detection is essential to prevent vision-threatening complications. This study ev...BACKGROUND: Diabetic retinopathy (DR) is a leading cause of preventable blindness among people with type 2 diabetes mellitus (T2DM). Early detection is essential to prevent vision-threatening complications. This study evaluated the long-term cost-effectiveness of CODMAP (Ophthalmologic Screening for Diabetes Mellitus in Primary Care (Cribado Oftalmológico en Diabetes Mellitus en Atención Primaria)), an optimised screening programme combining two-field non-mydriatic fundus photography (NMFP) and optical coherence tomography versus conventional single-field NMFP in a public healthcare setting. METHODS: A Markov model simulated DR progression over 50 years in a cohort of 7729 patients with T2DM. Eight health states reflected DR severity, with state-specific costs and quality-adjusted life years (QALYs) accrued annually. Analyses took the perspective of the Spanish National Health Service, applying a 3% annual discount rate. Incremental cost-effectiveness ratios, calculated as a ratio of means (ICERs), and net monetary benefits (NMBs) were estimated through 10 000 Monte Carlo simulations, at a €30 000 willingness-to-pay threshold. Deterministic and probabilistic sensitivity analyses, scenario analyses and bootstrap validation were performed. RESULTS: CODMAP accrued higher costs (€206.4 million vs €205.1 million) but more QALYs (128 691.7 vs 118 013.8), resulting in an ICER of €124.25/QALY. The probability of cost-effectiveness at €30 000/QALY was 74.3%, remaining stable in scenario analyses (75.7%-78.8%). Mean incremental NMB in the base case was €319.0 million. Cost and QALY variations were the main drivers of uncertainty. CONCLUSION: CODMAP offers greater long-term health gains at a favourable incremental cost per QALY in the Spanish context. However, the non-definitive probability of cost-effectiveness warrants cautious policy consideration, with attention to local infrastructure, implementation capacity and cost structures.
PURPOSE: Fundus fluorescein angiography (FFA) is an important tool in evaluating retinal vascular disease. In the era of optical coherence tomography angiography (OCTA), however, expert preferences regarding the comparat...PURPOSE: Fundus fluorescein angiography (FFA) is an important tool in evaluating retinal vascular disease. In the era of optical coherence tomography angiography (OCTA), however, expert preferences regarding the comparative utility of FFA and OCTA remain unclear. Additionally, despite FFA's widespread use, variability exists in the terminology used to describe angiographic findings. This study aimed to establish expert consensus on clinical indications for FFA versus OCTA and to provide consensus definitions of key angiographic terms. METHODS: Using a two-round modified Delphi process, 25 retinal subspecialists provided perspectives on the clinical indications for FFA in the assessment of a range of retinal vascular conditions. They also evaluated proposed definitions for FFA findings in retinal vascular diseases. Consensus was defined as ≥80% agreement and near consensus as 70%-79%. RESULTS: Experts agreed that FFA is preferable for the diagnosis of retinal vasculitis, ocular ischaemic syndrome and proliferative diabetic retinopathy, even when OCTA is available. Furthermore, FFA was the favoured imaging modality to guide laser photocoagulation in branch retinal vein occlusion. Conversely, FFA was considered non-essential in evaluating neovascular age-related macular degeneration and mild-to-moderate non-proliferative diabetic retinopathy. Finally, definitions were agreed on for seven FFA terms used in the evaluation of retinal vascular diseases. These were , , , , , and . CONCLUSION: This study presents contemporary perspectives on the clinical indications for FFA in an era in which OCT and OCTA are widely available. It also provides a lexicon for FFA reporting in retinal vascular diseases based on expert consensus.
PURPOSE: To evaluate the prognostic value of estimated tumour volume as a continuous biomarker in uveal melanoma and to compare its discriminative performance with the American Joint Committee on Cancer (AJCC) staging an...PURPOSE: To evaluate the prognostic value of estimated tumour volume as a continuous biomarker in uveal melanoma and to compare its discriminative performance with the American Joint Committee on Cancer (AJCC) staging and multiplex ligation-dependent probe amplification (MLPA)-based genetic risk classification. METHODS: This retrospective cohort study included consecutive patients with choroidal or ciliary body melanoma treated between 2016 and 2024 at a tertiary ocular oncology centre. Tumour volume (mm³) was calculated from baseline ultrasonography using an ellipsoidal (oval-base) approximation. The primary endpoint was metastasis-free survival (MFS). Survival was analysed using Kaplan-Meier methods and Cox regression. In the genetic subset, Firth-penalised Cox regression was applied. Prognostic discrimination was assessed using Harrell's concordance index (C-index) for AJCC stage, tumour volume, MLPA profile and combined models. RESULTS: A total of 237 patients were included (mean age 65.5±13.8 years; 54.0% female). Median follow-up was 61.0 months, and 64 patients developed metastasis during follow-up. Higher tumour volume was significantly associated with worse MFS, corresponding to an approximately 10% increase in hazard per 100 mm³ (HR 1.10; 95% CI 1.05 to 1.13; p<0.001). Continuous tumour volume showed higher discrimination than AJCC staging alone (C-index 0.73 vs 0.66). The highest discrimination was observed in models combining tumour volume and MLPA profile, with limited additional gain after including AJCC stage (C-index 0.79 in the MLPA subset). CONCLUSIONS: Continuous tumour volume provides more refined prognostic stratification in uveal melanoma and complements AJCC staging. Integrating volumetric, anatomic and molecular biomarkers improves risk assessment and may support individualised surveillance strategies.