Investigative Opthalmology & Visual Science

Age-related macular degeneration (AMD) is a leading cause of irreversible vision loss in ageing populations, with oxidative stress recognised as a key pathogenic driver. The dietary antioxidant and cytoprotectant, L-ergothioneine (ET), is avidly accumulated in many tissues, especially the eye. However its relationship to AMD has not been investigated. Here, we examined ETs distribution in ocular tissue and assessed circulating and intraocular ET levels in patients with neovascular AMD. Compared with ocularly-normal age-matched individuals, AMD patients exhibited significantly lower serum ET; elevated levels of ET metabolites, hercynine and ETSO, which may be generated by oxidative stress; and elevated levels of serum allantoin, a product of oxidative damage to urate in humans. Levels of ET in aqueous humour in AMD patients were marginally lower than cataractous patients who are already known to have significantly lower ET levels than healthy eyes. High ET levels were seen in human ocular tissues concentrating in regions vulnerable to oxidative injury, including the lens, retina, retinal pigment epithelium, and choroid, supporting a physiological protective role of ET in the eye. These findings identify the strong association between low ET levels and AMD, warranting further studies to determine whether ET supplementation can modify AMD risk or progression.

Primary open-angle glaucoma (POAG) disproportionately affects individuals of African ancestry, yet rare coding variation in this population remains understudied. To address this gap, we performed a multi-cohort exome-wide meta-analysis across POAAGG, PMBB, All of Us, and UK Biobank, including 4,815 POAG cases and 22,922 controls of genetically inferred African ancestry. Although no gene reached exome-wide significance, we identified several suggestive gene-level associations driven by rare variants (minor allele frequency [≤]0.1% or singletons),including signals in SRF, BLTP3A, METTL2A, and KRT10. Among these, SRF demonstrated the strongest association and was driven by rare missense variants with moderate effect sizes. Given its role in cytoskeletal organization and actin dynamics; processes central to trabecular meshwork function and intraocular pressure regulation SRF represents a biologically plausible candidate gene. Notably, these genes have not been previously highlighted in predominantly European ancestry POAG association studies, suggesting potential ancestry-specific rare variant contributions. Overall, our findings highlight the critical importance of investigating rare coding variation in POAG, in disproportionately affected populations to deepen understanding of POAG etiology and genetic risk.

ImportanceGenome-wide association studies have identified hundreds of common single nucleotide polymorphisms (SNPs) and small insertions/deletions (indels) associated with primary open-angle glaucoma (POAG) risk, though these variants have modest effect sizes and individually may have minor contributions to disease development. As whole-genome sequencing data is becoming more readily available, structural variants and other complex genomic features can be interrogated for contribution to disease risk. ObjectiveTest the association of structural variants in known glaucoma loci with disease risk. DesignCross-sectional study. SettingA multicenter cohort of individuals from the United States who contributed genomic and electronic health record data to the All of Us Research Program. ParticipantsPOAG case/control cohorts were generated in the All of Us Researcher Workbench using age (>40 for cases, >65 for controls) and ICD 9/10 diagnosis codes. Main Outcomes and MeasuresLogistic regression analyses adjusted for age, sex, and the top 10 principal components of ancestry were used to test association of structural variants within 500 kilobases of 309 known open-angle glaucoma risk loci. The significance threshold after Bonferroni correction was set at p<1.6x10-4. Results516 POAG cases and 18,716 controls of European ancestry from the All of Us v8 data release were included in the analysis. Mean age was 77.0 years among cases and 74.7 years among controls. Females comprised 45.7% of cases and 56.5% of controls. An 8,732 base pair deletion upstream of PITX2 (chr4:110680827-110689558) was associated with 7.3-fold higher odds of POAG (95% confidence interval: 2.9-18.5, p= 2.4x10-5, variant carrier frequency= 1.6% in cases and 0.25% in controls). Functional annotation identified multiple enhancers overlapping the deletion, suggesting that this structural variant likely impacts gene regulation and expression. Conclusion and RelevanceWhole genome sequencing data captures rare structural variants with large effect sizes that are missed by conventional SNP and indel genotyping approaches, enabling improved POAG risk stratification. These data also expand the phenotypic spectrum of structural variation in the PITX2 locus from childhood glaucoma to adult-onset disease, where age at diagnosis and clinical severity may be influenced by the extent of disrupted regulatory elements.

PurposeTo investigate the effects of morning and evening narrowband blue light exposure on axial length, and to examine the short-term effect of morning blue light combined with myopic defocus on axial length. MethodsFor objective 1, 18 individuals underwent 60 minutes of narrowband blue light exposure (460nm) in the morning (9:00-11:00AM) and evening (5:00-7:00PM) of the same day. The axial length values were normalized to the average of the morning and evening axial length values. For objective 2, 27 young adults were exposed to 60 minutes of narrowband blue light and broadband white light while wearing a +3.00 D lens over the right eye. Axial length was measured using Lenstar LS900. ResultsA significant reduction in axial length was observed after exposure to morning blue light compared to evening blue light (-10.0{+/-}3.96{micro}m vs.-0.67{+/-}3.30{micro}m; p=0.02), whereas no such effect was observed with broadband white light exposure (0.0{+/-}3.53 {micro}m vs. -2.50{+/-}4.23{micro}m, p=0.70). While the broadband white light exposure did not alter the normal diurnal variation in axial length (+2.35{+/-}1.82{micro}m vs.-6.25{+/-}2.21{micro}m, p=0.04), blue light diminished such a pattern (-4.12{+/-}1.72{micro}m vs. - 2.00{+/-}2.00{micro}m, p=0.48). The myopic defocus did not influence axial length under either narrowband blue or broadband white light conditions. ConclusionThe short-term narrowband blue light exposure led to a significant decrease in axial length in the morning than evening exposure, with a likely influence on the diurnal rhythm of axial length. Morning blue light exposure with lens-induced myopic defocus did not provide additional short-term modulation of axial length.

PurposeQuantitative metrics obtained from retinal fundus images (such as vessel length, tortuosity and other scale-dependent measures) are increasingly used as potential biomarkers for systemic diseases, including cardio- and neurovascular conditions. However, with the increasing prevalence of myopia and related axial growth, this study aims to evaluate if axial length scaling significantly alters the overall distributions of the inferred biomarkers when compared to biomarker data obtained without axial length scaling and if these effects can be corrected. Methods2,309 clinic visits from patients aged [&le;]21 years were analysed and extracted for axial-length scaling analysis (range) 20 to 28 mm). The retinal fundus photographs were automatically segmented using Automorph to extract biometric data, including vascular metrics. The parameters were further corrected for axial length using correction factors based on the Bennett-Littmann formula and true axial length. ResultsAxial length significantly influenced biometric parameters (vessel metrics) derived from fundus photography. The magnitude of error in diameter and length of blood vessels was approximately 4-5% for each 1 mm deviation from the reference axial length of 24 mm, whereas the error in vessel area was approximately 9-10% per 1 mm, consistent with the geometric expectation that area scales with the square of linear dimensions. The scaling corrections for different axial lengths are presented. ConclusionsAxial-length-related magnification introduces systematic bias into retinal vascular metrics from fundus photographs. Bennett-Littmann correction using true axial length reduces these errors and should be adopted in quantitative fundus imaging and Al biomarker development.

PurposeThere is a need for novel therapies for diabetic retinopathy (DR) because existing therapies treat only certain features of DR and do not work optimally for all patients. While proteomic studies provide insight into disease pathobiology, they are often limited to small sample sizes due to high costs, limiting their generalizability and reproducibility. Moreover, they often yield lists of tens to hundreds of proteins with differential expression, making it difficult to prioritize the most biologically relevant biomarkers beyond using arbitrary fold-change and false-detection rate cutoffs. Here, we applied a two-stage multimodal AI approach: first, we integrated EHR and proteomics data to rationally prioritize candidate protein biomarkers and, next, validated these biomarkers in an independent cohort. These protein biomarkers of DR are rooted in the EHR data and thereby more likely to be biological drivers of disease. MethodsWe obtained EHR data from a large number of patients with and without DR (N=319,997) from the STARR-OMOP database and obtained aqueous humor liquid biopsies from a subset of these patients (N=101) for high-resolution proteomic profiling. We developed Clinical and Omics Multi-Modal Analysis Enhanced with Transfer Learning (COMET) to perform integrated analysis of proteomics and all available EHR data to identify protein biomarkers of DR. The model was trained in two phases: first, it was pretrained using patients with EHR data alone (N=319,896), and then, it was fine tuned using patients with both EHR and proteomics data (N=101), allowing it to learn both clinical and molecular features associated with DR. Findings from COMET were then validated with liquid biopsies from an independent, validation cohort (N=164). Resultst-distributed stochastic neighbor embedding (t-SNE) analysis of EHR and proteomics data identified proteins clustering with related EHR features. Levels of STX3 and NOTCH2, proteins involved in retinal function, were correlated with a diagnosis of macular edema, a record of a visual field exam, and a prescription for latanoprost, highlighting protein-EHR alignment. The pretrained, multimodal COMET model was superior (AUROC=0.98, AUPRC=0.91) compared to models generated using either EHR or proteomics data alone or without pretraining (AUROC: 0.76 to 0.92; AUPRC: 0.47 to 0.74). The proteins SERPINE1, QPCT, AKR1C2, IL2RB, and SRSF6 were prioritized by the COMET model compared to the models without pretraining, supporting their potential role in DR pathobiology, and were subsequently validated in an independent cohort. ConclusionWe used multimodal AI to prioritize protein biomarkers of DR that are most strongly linked to EHR elements, as well as identifying other protein biomarkers associated with disease features like diabetic macular edema. These findings serve as a foundation for future mechanistic studies and highlight the synergistic value of using multimodal AI to fuse EHR and proteomics data for enhanced proteomics analysis.

IntroductionAnti-vascular endothelial growth factor (anti-VEGF) therapies are standards of care for vision-threatening retinal diseases. This retrospective observational study describes demographics, utilization, best recorded visual acuity (BRVA), and safety among eyes with neovascular age-related macular degeneration (nAMD), diabetic retinopathy (DR), diabetic macular edema (DME), or retinal vein occlusion (RVO) treated with the biosimilar aflibercept-ayyh (PAVBLU(R)) in routine clinical practice. MethodsElectronic medical records from the Retina Consultants of America database of patients receiving aflibercept-ayyh (12/1/2024-10/31/2025) were analyzed, focusing on eyes with [&ge;]84 days of follow-up. The index date was the first documented aflibercept-ayyh injection. Postindex data were used to assess treatment patterns, BRVA (Wilcoxon signed rank test), and adverse events of special interest (AESIs). ResultsA total of 1,000 consecutive eyes from 989 patients received 3,730 injections of aflibercept-ayyh; most (91%) switched from prior anti-VEGF therapy and 9% were anti-VEGF treatment-naive. Disease distribution was 58% nAMD, 19% RVO, 16% DME, and 7% DR. Among switchers, median (IQR) number of prior injections was 21 (8-46). Median (IQR) follow-up was 6.0 months (4.6-7.1). Median (IQR) number of aflibercept-ayyh injections per eye was 4 (3-5). Among eyes with [&ge;]84 days of follow-up (n=889), mean BRVA expressed as logarithm of minimum angle of resolution (logMAR) remained stable for switchers (0.4 to 0.4; P=0.96) and improved from baseline in anti-VEGF-naive eyes (0.5 to 0.4; P<0.01). Confirmed AESIs included iritis (n=2; 0.05% of injections), with no events of vitreous cells, endophthalmitis, retinal detachment, retinal vasculitis, or vitreous hemorrhage. ConclusionIn this descriptive real-world analysis, aflibercept-ayyh was associated with stable visual acuity in previously treated eyes and vision improvement in treatment-naive eyes, with no new or unexpected safety findings, consistent with expectations for aflibercept. These findings add real-world experience to preexisting evidence demonstrating no clinically meaningful differences between aflibercept-ayyh (PAVBLU(R)) and reference aflibercept (EYLEA(R)). KEY SUMMARY POINTSO_ST_ABSWhy carry out this study?C_ST_ABSO_LIThe anti-vascular endothelial growth factor (VEGF) drug aflibercept, approved in 2011 and marketed in the United States as EYLEA(R),* has demonstrated efficacy in treating retinal diseases such as neovascular age-related macular degeneration (nAMD), diabetic retinopathy (DR), diabetic macular edema (DME), or retinal vein occlusion (RVO) and is a standard of care for these disorders. C_LIO_LIAflibercept-ayyh is a biosimilar to aflibercept that has demonstrated comparable efficacy and safety in the treatment of nAMD in a randomized controlled clinical trial. C_LIO_LIThis study describes the real-world use patterns, vision outcomes, and safety of aflibercept-ayyh in clinical settings in the United States for the treatment of nAMD, DR, DME, and RVO. C_LI What was learned from the study?O_LIIn this real-world study of 1,000 consecutive eyes treated with the biosimilar aflibercept-ayyh in patients with retinal diseases, we observed no new safety concerns and that aflibercept-ayyh maintained visual acuity in eyes switching anti-VEGF agents and improved vision in anti-VEGF-naive eyes, consistent with expected responses to aflibercept. C_LIO_LIThese findings support aflibercept-ayyh as a suitable treatment option when anti-VEGF therapy is indicated. *EYLEA(R) is a registered trademark of Regeneron Pharmaceuticals, Inc. PAVBLU(R) is a registered trademark of Amgen Inc. C_LI

BackgroundDiabetes mellitus (DM) remains a major global health challenge and is associated with vision-threatening complications, including diabetic macular edema (DME), a leading cause of visual impairment. Dyslipidemia has been implicated in the development of macular edema through mechanisms involving vascular permeability, endothelial dysfunction, and chronic inflammation. However, evidence regarding the relationship between lipid abnormalities and macular edema remains inconsistent across studies. AimThis study aimed to evaluate the association between abnormal lipid profiles and diabetic macular edema among patients with type 2 diabetes mellitus attending Kilimanjaro Christian Medical Centre (KCMC). MethodsA hospital-based analytical cross-sectional study was conducted among 296 diabetic outpatients at KCMC. Participants underwent comprehensive ophthalmic evaluation including fundoscopy and imaging with optical coherence tomography (OCT) for assessment of macular edema. Blood samples were collected for biochemical lipid analysis. Data were cleaned and analyzed using STATA version 17. ResultsDiabetic macular edema was identified in 56.4% (167/296) of participants. Abnormal lipid parameters were common, with elevated total cholesterol observed in 48.6%, triglycerides in 43.6%, low-density lipoprotein (LDL) in 36.1%, and reduced high-density lipoprotein (HDL) in 38.9% of patients. Elevated total cholesterol, triglycerides, and LDL levels showed significant associations with macular edema (p < 0.05). After multivariable adjustment, serum triglycerides remained independently associated with macular edema (p = 0.002). ConclusionDyslipidemia demonstrated a significant association with diabetic macular edema, with serum triglycerides emerging as an independent predictor. These findings highlight the importance of lipid monitoring, lifestyle modification, and strengthened screening strategies in reducing the burden of vision-threatening diabetic complications.

PurposeTo systematically evaluate ocular biometric and systemic laboratory factors associated with cataract in highly myopic eyes and to characterize potential nonlinear associations using an interpretable machine learning approach, thereby providing deeper mechanistic insights into the pathogenesis of highly myopic cataract. DesignA cross-sectional study encompassed 770 eyes of 594 patients with high myopia from Eye & ENT Hospital of Fudan University. SubjectsThe non-cataract control group included 458 eyes while the cataract group contained 312 eyes. MethodsDemographic traits, ocular biometric and systemic laboratory factors were gathered while features with over 30% of missing data were excluded. Composite indices were obtained through calculation. Multiple machine learning models were compared to investigate the association between features and highly myopic cataract, and the random forest (RF) model was chosen and fine-tuned. Feature selection was carried out by means of Shapley additive explanations (SHAP) and non-linear relationships were probed using SHAP dependence diagrams and confirmed with partial dependence plots. Main Outcome Measures(1) The Area Under the Curve (AUC) and other metrics of multiple machine learning models; (2) Top feature importance of the final simplified RF model; (3) Overall trends between features and highly myopic cataract; (4) Potential inflection points of top continuous features. ResultsA simplified fine-tuned RF model with 17 features reached stable discriminative performance, with a mean AUC of 0.762 (95%CI: [0.731, 0.794]) among 10 independent testing sets. Age and axial length (AL) turned out to be the most influential features which had non-linear relationships highly myopic cataract, with an inflection point seen around 65.75 (95%CI: [63.72, 67.79]) years for age and 30.55 (95% CI: [29.22, 31.88]) mm for axial length respectively, while the ratio of anterior chamber depth to axial length (ACD/AL) was associated with highly-myopic cataract in a U-shape. Ocular biometric factors were more strongly related to highly myopic cataract than systemic laboratory factors. ConclusionsOcular biometric factors, especially age, AL, and composite indices like ACD/AL, have strong and non-linear connections with highly myopic cataract. These results emphasize the significance of ocular structural arrangement in cataract within highly myopic eyes and indicate that interpretable data-driven methods could offer clinically relevant understandings regarding its phenotypic description.

BackgroundThe visual field (VF) test results of many eyes with glaucoma progress despite treatment. This suggests that some eyes are either untreated or that the management of intraocular pressure (IOP) does not influence the outcome. In this work, we explore whether future VF parameters can be predicted from a baseline optical coherence retinal nerve fibre layer (OCT-RNFL) scan using a deep learning model. MethodsThe model was developed using 1792 eyes from 1610 patients, and externally validated on 151 eyes from a second centre using the same Zeiss Cirrus machine and 281 eyes from a third centre using scans obtained from a different (Heidelberg Spectralis) machine. The Vision Transformers (ViT)-based regression model was trained on baseline OCT-RNFL scans to predict three key VF indices (follow-up interval: 4.74 {+/-} 2.59 years). Model performance was evaluated using Mean Absolute Error (MAE) and Root Mean Square Error (RMSE), with 95% confidence intervals (CI). ResultsThe model achieved an overall MAE of 2.07 (95% CI: 1.91-2.22) and RMSE of 2.87 (95% CI: 2.60-3.14) on the internal validation set. On external validation, the model showed comparable performance with an MAE of 2.07 (95% CI: 1.8-2.35) for the external validation (Zeiss OCT) cohort and 2.11 (95% CI: 1.93-2.31) for the external validation (Heidelberg OCT) cohort. Saliency maps revealed that the inner and outer RNFL layers were key structures in driving the models predictions. ConclusionsOur ViT-based regression model effectively predicts key VF indices objectively from a single OCT-RNFL scan, with strong performance across two OCT devices, offering a novel tool for predicting glaucoma progression.

Regular optical coherence tomography (OCT) monitoring is essential for early detection of retinal disease and timely intervention, but frequent clinicbased imaging burdens patients and healthcare systems. Home-based OCT enables continuous monitoring and reduces clinic visits; however, compact optics and patient-operated acquisition introduce noise, reduced resolution, motion blur, and artifacts that limit clinical reliability and diagnostic confidence. To model home-based OCT acquisition, we employ simulated data reflecting images from Siloton, a compact home-based OCT device. Clinically realistic noise and acquisition artifacts were applied to high-quality OCT images using Silotons simulation software, generating near-real patient-operated scans. Building on this dataset, we propose HAGAN, a Hybrid Attention Generative Adversarial Network developed through a progressive strategy, evolving from a baseline U-Net to an adversarial framework with hybrid attention. The best-performing U-Net architecture, EfficientNet-B1, identified through evaluation and ablation studies, is adopted as the generator. The generator incorporates attention gates at its skip connections and self-attention modules within the decoder, and is paired with a VGG19-based discriminator to form the HAGAN architecture. The model is trained using a multiobjective loss combining pixel-wise, structural, perceptual, edge-preserving, and adversarial components. Experiments on simulated home-based OCT data demonstrate that HAGAN consistently outperforms baseline and state-of-the-art models across standard enhancement metrics and a clinically relevant retinal layer segmentation downstream task, improving visual quality and preservation of diagnostically meaningful anatomical structures. These findings support the potential of HAGAN for reliable enhancement in future home-based OCT platforms, enabling remote retinal monitoring and reducing reliance on in-clinic imaging and routine hospital visits. HighlightsO_LIEnhancing the quality of home-based OCT images to support remote retinal monitoring and reduce the need for frequent referrals to clinical imaging centers C_LIO_LIProposing HAGAN, a hybrid attention generative adversarial network for enhancing OCT images acquired using the Siloton home-based OCT device C_LIO_LIHybrid attention design combining attention gates and self-attention to preserve fine retinal details and global anatomical consistency C_LIO_LIAdversarial learning framework improving perceptual realism and preservation of diagnostically relevant retinal structures in low-quality homeacquired OCT images C_LIO_LIProgressive model development from baseline U-Net to hybrid attention GAN, demonstrating systematic and measurable performance improvements C_LIO_LIClinical relevance validated through downstream retinal layer segmentation, confirming preservation of diagnostically important structures C_LI

PurposeTimely detection of disease activity in chronic retinal diseases improves visual outcomes but is limited by the lack of validated systems for continuous monitoring and care management. We evaluated the real-world performance of an integrated remote physiologic monitoring and principal care management program (RemoniHealth(R)) using a self-administered multimodal retinal function test (Macustat(R)) for home monitoring. MethodsThis single-arm real-world intervention study was conducted across 33 retina practices. A total of 2,216 adults with chronic retinal diseases performed weekly home retinal function testing with integrated care management support. Primary endpoints included the annualized rate of disease progression detection, time to intervention after first flag, true positive rate, and patient adherence. Descriptive statistics and data analyses were analyzed using chi-square tests and Clopper-Pearson confidence intervals. ResultsParticipants contributed 82,644 encounters and 16,805 patient-months of monitoring. The program generated 241 alerts, including 101 Macustat flags and 135 care management prompts. Among 73 adjudicated flags, 56 were true positives and 17 false positives (PPV 76.7%). The annualized detection rate was 4 per 100 patient-years. Of confirmed events, 93% led to intravitreal injection or other major management change. Mean adherence was 72.1%, and patients with [&ge;]80% adherence had higher odds of true positivity. DiscussionThis RPM-PCM model achieved high engagement and meaningful detection of asymptomatic progression between visits, supporting the value of home monitoring for timely intervention. Translational RelevanceThese findings support scalable integration of home vision testing and care management into routine retinal practice to enable earlier intervention and improved continuity of care.

Early screening for glaucoma and diabetic retinopathy (DR) is critical to prevent irreversible vision loss, yet remains inaccessible to many underserved populations. However, AI models trained on hospital-grade fundus images often generalize poorly to low-cost images acquired with portable devices such as smartphones. We proposed CausalFund, a causality-inspired learning framework for training AI models that enable reliable low-resource screening from easily acquired non-clinical images. CausalFund disentangles disease-relevant retinal features from spurious image factors to achieve domain-generalizable screening across clinical and non-clinical settings. We integrated CausalFund with seven deep learning backbones for glaucoma and DR screening from portable-device fundus images, including lightweight architectures suitable for on-device deployment. Across diverse experimental settings and image quality conditions, CausalFund consistently improved AUC and achieved a more favorable sensitivity-specificity trade-off than conventional deep learning baselines. As a model-agnostic framework, CausalFund could be extended to other diseases and low-resourced scenarios characterized by degraded or non-standard imaging.

PurposeFuchs endothelial corneal dystrophy (FECD) is a common corneal disease and a leading indication for endothelial keratoplasty (EK). Although CTG18.1 repeat expansion is a major genetic risk factor, the contribution of polygenic background to disease progression remains unclear. We evaluated whether combining CTG18.1 expansion status with a FECD-specific polygenic risk score (PRS) enables genomic prediction of progression to EK. MethodsWe retrospectively analysed 589 individuals with FECD from two European centers, with replication in an independent cohort of 185 individuals. Association of CTG18.1 expansion ([&ge;]50 repeats) and PRS with time to EK were evaluated using Cox models adjusted for sex and ancestry. ResultsExpansion-positive status was associated with earlier EK (HR 2.30; 95% CI 1.62- 3.26; P<.001). Addition of PRS improved prediction (C-index 0.614 vs 0.602; P=.014). Each 1-SD increase in PRS was associated with earlier EK (HR 1.16; 95% CI 1.03-1.30; P=.015), with replication in the validation cohort (HR 1.42; 95% CI 1.15-1.75; P=.001). ConclusionIntegration of monogenic and polygenic risk enables genomic prediction of FECD progression, supporting clinical genomic risk stratification to inform individualized monitoring and timing of intervention.

ObjectiveTo identify risk loci for Fuchs endothelial corneal dystrophy (FECD) and improve a genetic risk prediction model. DesignGenome-wide association study (GWAS), polygenic risk score (PRS) construction, and TCF4 CTG18.1 short tandem repeat (STR) length inference. ParticipantsThe study included 7,316 Europeans (EUR) with FECD or related corneal dystrophy phenotypes and 1,588,467 controls from the UK Biobank, All of Us, FinnGen, and the Million Veteran Program. Two independent EUR FECD cohorts were used for PRS validation (1,851/2,679 cases/controls and 124/257 cases/controls). African (AFR) ancestry analyses included 455 cases and 121,154 controls to build PRS. A subset of All of Us participants was used for joint PRS and STR modelling. MethodsGWAS meta-analyses were performed using FECD diagnoses or corneal dystrophy proxies where necessary, with validity assessed via genetic correlation. Risk loci were identified, and ancestry-specific PRSs were constructed using SBayesRC. PRS performance was evaluated across ancestries with and without TCF4 STR data. Main OutcomeWe identified novel loci for corneal dystrophy and constructed PRS-based and STR-based prediction models. ResultsThe GWAS meta-analysis identified 24 risk loci associated with corneal dystrophy, including 12 novel loci, doubling previous FECD studies. The optimised PRS outperformed existing models in two independent FECD validation cohorts (AUC = 0.83, 95% CI: 0.82-0.84; DeLongs P = 7.04 x 10-19), with individuals in the top PRS decile showing 14-fold and 19-fold increased risk in the two validation sets, respectively In All of Us, STR expansion (>40 repeats) was the key predictor of FECD risk, yielding excellent discrimination (AUC = 0.89; OR = 54) with minimal improvement from PRS. Consistent with this, STR expansion remained the primary driver of risk across ancestries, while PRS provided modest independent value for broader corneal dystrophy phenotypes in EUR and admixed American populations. Among participants without large STR expansion, overall predictive performance was modest; PRS was the only significant genetic contributor (OR = 1.37) for broader corneal dystrophy in Europeans, whereas analyses in FECD non-expansion carriers were underpowered. ConclusionsThese findings refine the genetic architecture of FECD, enhance risk prediction, and support a tiered strategy integrating STR expansion testing with PRS. Key PointsO_ST_ABSQuestionC_ST_ABSCan polygenic risk scores (PRS), alone or combined with TCF4 CTG18.1 short tandem repeat (STR) length, improve genetic risk prediction for Fuchs endothelial corneal dystrophy (FECD)? FindingsIn this GWAS meta-analysis of 7,316 cases and 1,588,467 controls, PRS showed strong predictive performance in validation cohorts lacking STR data. When STR length was available, it was the main predictor of FECD risk with limited additional contribution from PRS. Among non-expansion STR carriers, PRS helped stratify risk for broader corneal dystrophy in Europeans. MeaningPRS provide a practical, complementary approach for FECD risk prediction, particularly when STR data are unavailable.

BackgroundTuberculosis (TB) incidence peaks in women during their reproductive years and is a leading cause of maternal mortality. Pregnant women with TB have a high risk of failure to initiate TB treatment and poor TB treatment and pregnancy outcomes. We determined the time to treatment initiation in pregnant women diagnosed with TB in a routine programmatic setting. MethodsUsing routine linked electronic data, we identified women 15-45 years of age with laboratory-confirmed and/or clinically diagnosed TB, October 2018-December 2020, in two high-burden sub-districts in Cape Town, South Africa. We compared demographic and clinical characteristics in women with TB by pregnancy status, used time-to-event analysis to determine the time from TB diagnosis to initiation of antituberculosis treatment and Cox regression to assess determinants of treatment initiation. ResultsOf 5,459 women diagnosed with TB, 292 (5.3%) were pregnant. The median age for pregnant women was 28.6 years (interquartile range [IQR]: 23.7-33.7) and non-pregnant women 31 years (IQR:25.2-36.5). HIV prevalence was similar in pregnant (177/292; 60.6%) vs non-pregnant (3200/5167; 61.9%) women. Median time to treatment initiation was two days for pregnant and non-pregnant women. Most women initiated treatment within the first month after their TB diagnosis, after which the rate plateaued in both groups. Time to treatment initiation over 6 months was statistically different (Kaplan Meier Log-rank test, p = 0.0064) with pregnant women lagging behind non-pregnant women. ConclusionsMore than 5% of women diagnosed with TB were pregnant at the time of TB diagnosis. While pregnant women with TB were appropriately initiated on treatment, almost 15% were never started on treatment and there were delays in treatment initiation. While strategic interventions to prioritise early treatment initiation are needed, there should be a specific focus on pregnant women who have not initiated treatment within one month after TB diagnosis.

IntroductionFibrosis can affect organs throughout the body and is present in a wide range of diseases. Recent research has suggested that there could be shared biological mechanisms that lead to fibrosis in different organs. MethodsWe performed genome-wide association studies using UK Biobank for fibrosis in 12 different organ-systems and meta-analysed results with previously published studies of fibrotic diseases. We considered genetic associations that colocalised across [&ge;]3 organs as those likely to be involved in general fibrotic mechanisms and also identified novel genetic variants not previously reported as associated with fibrosis. Genetic correlation of fibrosis between organs was calculated using linkage disequilibrium score regression (LDSC). Discovery analyses were performed using European ancestry individuals and results were tested further in African, South Asian and East Asian ancestry groups. ResultsWe identified eight genetic loci that colocalised across three or more organs. One of these signals, located near the SH2B3 and ATXN2 genes, showed evidence of a shared causal variant for fibrosis across five organs. We also identified two novel fibrotic associations, one implicating alternative splicing of TFCP2L1 for urinary fibrosis and another implicating a missense variant in FAM180A for intestinal-pancreatic fibrosis. We observed significant genetic correlations for all organs, particularly for liver and skeletal fibrosis. ConclusionWe found evidence of shared genetic associations for fibrosis across organs, both at individual genetic loci and genome-wide. This highlights specific genes that may contribute to fibrosis across organs and diseases, which may facilitate the development of new therapies.

ObjectiveHbA1c thresholds used to define dysglycemia in autoantibody-positive individuals at risk for type 1 diabetes do not account for age-related increases in HbA1c and may overestimate progression risk in adults. We evaluated whether age-adjusted HbA1c or a higher HbA1c threshold improves risk stratification across age groups. Research Design and MethodsWe analyzed 5,024 autoantibody-positive relatives (3,720 children and 1,304 adults) participating in the TrialNet Pathway to Prevention study. Age-related HbA1c effects were modelled using 6,273 adults from the population-based Exeter 10,000 cohort. Progression risk was compared using the standard dysglycemia threshold (HbA1c [&ge;] 5.7% [39 mmol/mol]), age-adjusted HbA1c, and an alternative threshold of HbA1c [&ge;]6.0% (42 mmol/mol). ResultsUsing HbA1c [&ge;] 5.7%, children had higher 1-year progression risk than adults among single autoantibody-positive participants (38% [95% CI 28, 47] vs. 13% [7.2, 19]) and multiple autoantibody-positive participants (55% [49, 60] vs. 38% [27, 47]; both p<0.001). Age adjustment reduced these differences; progression risk was similar among single autoantibody-positive participants (38% [28, 47] vs. 27% [13, 39]; p=0.32), with attenuated differences among multiple autoantibody-positive participants. An HbA1c threshold [&ge;]6.0% yielded comparable progression risk between adults and children across autoantibody subgroups. In post hoc analyses, adults aged <30 years had progression risk similar to children (p=0.1). ConclusionsAge-related variation in HbA1c influences dysglycemia classification in adults at risk for type 1 diabetes. Age-adjusted HbA1c or a higher HbA1c threshold ([&ge;]6.0% [42 mmol/mol]) in adults [&ge;]30 years identifies individuals with progression risk comparable to children and may improve age-specific risk stratification in prevention seungs.

Non-alcoholic fatty liver disease (NAFLD) is a globally prevalent hepatic condition caused by the buildup of fat in the liver. It is frequently associated with metabolic comorbidities such as hypertension, cardiovascular disease (CVD), and prediabetes. However, early detection remains challenging due to the asymptomatic progression, and existing primary diagnostic methods, such as imaging or liver biopsy, are often expensive and inaccessible in rural areas. This study proposes a two-stage, interpretable machine learning pipeline for the non-invasive and cost-effective prediction of NAFLD and its key comorbidities using routine clinical parameters. The NAFLD prediction model was developed using the XGBoost algorithm, trained on a hybrid dataset that combines real patient data with rule-based synthetic data generated by simulating clinically plausible cases. Upon NAFLD-positive prediction, three separate XGB models, trained on data labelled based on thresholds, assess individual risks for hypertension, cardiovascular disease, and prediabetes. Explainability is obtained using SHAP (SHapley Additive exPlanations), which provides insight into feature relevance, while biomarker radar plots help in the visual interpretation of comorbidities. A user-friendly Streamlit interface enables real-time interaction with the tool for potential clinical application. The NAFLD model demonstrated robust performance, while the models used for predicting comorbidities achieved perfect performance, which may be a reflection of the limited dataset size used in the second stage. This work underscores the potential of AI-driven tools in NAFLD diagnosis, particularly when combined with explainable AI methods.

ObjectiveTo quantify trends in annual mean healthcare costs per person living with HIV from 2003 to 2018 from a publicly funded healthcare system perspective. DesignWe conducted a retrospective population-based study using administrative health data in Ontario, Canada, including 25,842 people living with HIV diagnosed and entering care between 1992 and 2018. A nested cohort from the Ontario HIV Treatment Network Cohort Study (n=3,516) provided additional HIV-related characteristics. MethodsAnnual mean healthcare costs per person were estimated using a validated costing algorithm and inflated to 2025 Canadian dollars. Trends were examined overall and stratified by sociodemographic factors (age, sex, rurality, neighbourhood income, immigration status) and year of entry into HIV care. Within the nested cohort, trends were stratified by nadir CD4 count and any antiretroviral therapy use since diagnosis. ResultsAnnual mean cost per person increased from $11,963 in 2003 to $16,721 in 2018. Medication costs remained the largest cost component throughout (47.4-61.7%) and closely mirrored overall trends. Higher annual mean costs were consistently observed among individuals diagnosed at older ages, lower-income neighbourhood residents, long-term Ontario residents (Canadian-born or immigrated before 1985), and individuals with nadir CD4<200cells/{micro}L. ConclusionMedication expenditures continue to drive healthcare costs for people living with HIV. Cost containing strategies, including expanded generic substitution and strengthened price negotiation, may reduce costs without compromising outcomes. Persistent cost disparities highlight the need to address delayed treatment initiation and broader social determinants shaping HIV treatment access and sustained engagement in care.