BACKGROUND:Aspirin prevents preterm preeclampsia in high-risk pregnancies, but the extent to which its effectiveness depends on adherence and baseline risk remains uncertain. OBJECTIVE:To evaluate how first-trimester preeclampsia risk and aspirin adherence jointly influence aspirin's preventive effect, and to assess the potential benefits and harms of targeted versus universal prophylaxis. STUDY DESIGN:We combined data from the Aspirin for Evidence-based Preeclampsia Prevention and the Screening Program for Preeclampsia cohorts, including singleton pregnancies delivering at ≥24 weeks. All pregnancies were screened for preterm preeclampsia at 11+0 to 13+6 weeks using the Fetal Medicine Foundation competing risks model, which combines maternal characteristics, mean arterial pressure, uterine artery pulsatility index, and serum placental growth factor. Aspirin adherence and baseline risk distributions were modeled using Monte Carlo simulations to estimate relative risk, absolute risk reduction, and number needed to treat under 4 adherence scenarios: full (100%), trial-based (intention-to-treat), 50%, and variable adherence positively correlated with predicted risk (ρ=0.4). Decision-curve analysis assessed net benefit across risk thresholds. RESULTS:Simulations informed by 51,024 pregnancies indicated that aspirin's preventive effect varied substantially with adherence. In fixed-adherence scenarios, the strongest effect occurred with full adherence (relative risk 0.25, 95% confidence interval 0.09-0.66) and the weakest with 50% adherence (relative risk 0.70, 95% confidence interval 0.58-0.98). Under variable adherence, relative risk decreased nonlinearly with baseline risk, approaching the per-protocol effect in high-risk women but near null in low-risk women. Absolute risk reduction and number needed to treat were highly dependent on predicted risk: at 1 in 50 to 1 in 100 risks, number needed to treat ranged from 73 to 146 with high adherence and 161 to 321 with 50% adherence, whereas at lower risks, numbers needed to treat exceeded several thousand even with high adherence. Decision-curve analysis indicated that targeted prophylaxis using the Fetal Medicine Foundation model provided greater net benefit than universal treatment, primarily by avoiding unnecessary interventions. CONCLUSION:Aspirin is highly effective for preventing preterm preeclampsia in women at increased risk, but its effect depends on predicted risk and adherence. The absolute benefit of treatment is negligible in low-risk populations. A targeted screen-and-treat strategy using the Fetal Medicine Foundation model maximizes clinical benefit while minimizing unnecessary treatment.

PURPOSE:To evaluate the efficacy of switching from bevacizumab to ranibizumab or aflibercept in neovascular age-related macular degeneration (nAMD) using automated volumetric retinal fluid analysis. METHODS:This retrospective study included 186 eyes with nAMD showing inadequate response after ≥3 bevacizumab injections. Patients were switched to a second-line anti-VEGF agent and categorized as early (3-5 injections) or late (≥6 injections) switchers. Optical coherence tomography (OCT) scans were analyzed using the Notal-OCT Analyzer (NOA), an AI-based algorithm for retinal fluid quantification. Parameters included total retinal fluid (TRF), intraretinal fluid (IRF) volume, subretinal fluid (SRF) volume, pigment epithelial detachment (PED) volume, central subfield thickness (CST), and visual acuity (VA), assessed at baseline, after bevacizumab, and after three injections of the new agent. RESULTS:Early switchers showed an increase in TRF from 105 nL to 158 nL after bevacizumab, then a reduction to 20 nL postswitch ( P < 0.001); PED volume decreased postswitch ( P = 0.010). Late switchers showed fluid reductions both after bevacizumab and postswitch (TRF: P < 0.001; PED: P = 0.007). CST significantly improved only in early switchers ( P = 0.004), while VA changes were not statistically significant. CONCLUSION:Automated fluid analysis reliably quantifies treatment response in nAMD. Early switchers showed worsening after bevacizumab, while late switchers had partial responses. These findings highlight the potential of automated analysis to support more individualized, data-driven treatment strategies. Prospective studies are required to confirm impact on patient outcomes.

Background:Ocular hypertension, that is intraocular pressure > 21 mmHg, is a risk factor for glaucoma. A glaucoma risk predictor, the Ocular Hypertension Study-European Glaucoma Prevention Study model, is available. Objectives:(1) To validate and update the Ocular Hypertension Study-European Glaucoma Prevention Study risk prediction model in a United Kingdom population; (2) to assess the relative efficiency of alternative monitoring pathways according to glaucoma risk; (3) to determine the clinical and cost-effectiveness of treating people with ocular hypertension with intraocular pressure of 22 or 23 mmHg and (4) to elicit patient preferences for monitoring. Design:(1) Retrospective data analysis of electronic medical records of ocular hypertension patients attending hospital eye services. The influence of the Ocular Hypertension Study-European Glaucoma Prevention Study predictors and additional ocular and systematic factors was explored. Validation: the Ocular Hypertension Study-European Glaucoma Prevention Study prediction model was applied. Update: the model was refitted by re-estimating baseline hazard and regression coefficients. (2, 3) Predictor versus standard care, with deterministic and probabilistic sensitivity analyses. Subgroup analysis for people with 22-23 mmHg intraocular pressure. (4) Discrete choice experiment. Setting and participants:People with intraocular pressure 22-32 mmHg in either eye, at least four visual field tests, 5 years of follow-up, no significant ocular comorbidities. Data sourced from secondary clinical settings. Main outcome measures:Discriminative ability (c-index) and calibration (calibration slope) to predict conversion to glaucoma in 5 years. Quality-adjusted life-years, incremental cost-effectiveness ratio, preferences. Data sources:Electronic medical records of 10 hospitals in England. Results:(1) Of 9030 patients with ocular hypertension who fitted the inclusion criteria 1530 (16.9%) converted to glaucoma. The Ocular Hypertension Study-European Glaucoma Prevention Study model provided a pooled c-index of 0.61 (95% confidence interval: 0.60 to 0.63). The updated model had a pooled c-index of 0.67 (0.51 to 0.84). (2) In the economic model almost all (99%) patients were treated in the risk predictor strategy, and less than half (47%) in the standard care strategy. The risk predictor strategy produced higher costs, but also higher quality-adjusted life-years and is likely to be cost-effective compared with standard care. (3) Patients with ocular hypertension and intraocular pressure 22-23 mmHg had similar risk of conversion to the rest of the cohort. A treat-all strategy may not be cost-effective. (4) Three hundred and sixty patients were recruited from four NHS hospitals. Almost all respondents (92%) had experienced face-to-face monitoring at a hospital; fewer respondents had experienced virtual clinics (47%) or community optometrist monitoring (43%). Most patients preferred hospital-based monitoring services by health professionals rather than community-based by optometrists but patients with prior experience of community optometrist monitoring preferred it. Patients preferred options associated with lower risk of conversion and lower costs. Limitations:Insufficient data to evaluate influence of ethnicity or ocular factors such as optic disc and retinal anatomy. Conclusions:We validated the Ocular Hypertension Study-European Glaucoma Prevention Study predictor model in a large population with ocular hypertension achieving modest improvements. The use of a risk prediction tool is likely to be cost-effective. Reducing the risk of conversion was the most important preference for patients with ocular hypertension. Future work:Future work should address the influence of genetic or other ocular factors in disease progression, evaluation of effectiveness and cost-effectiveness of different models of eye care, and on how to avoid late glaucoma presentation. Funding:This synopsis presents independent research funded by the National Institute for Health and Care Research (NIHR) Health Technology Assessment programme as award number NIHR131808.

Purpose:To evaluate retinal and choroidal vascular alterations in multiple sclerosis (MS) using multimodal imaging and determine their association with disability and disease progression. Design:Prospective, observational, single center, cross-sectional, case-control study. Participants:Sixteen MS and 25 control participants. Methods:Multimodal retinal imaging, including color fundus photography, ultra-widefield imaging, OCT, and OCT angiography, were performed. Retinal vascular parameters (RVPs) and choroidal vascular parameters (CVPs) were compared between control eyes and eyes with (eyes with a history of optic neuritis [MSON]) and without optic neuritis (ON) (eyes without a history of ON [MSnON]) using regression models. Associations of RVPs and CVPs with Expanded Disability Status Scale (EDSS) scores and annual EDSS progression rate were assessed. Subanalysis compared RVPs and CVPs between MSnON and MSON eyes using Wilcoxon rank-sum tests. Main Outcome Measures:Differences in RVPs and CVPs between groups and associations with EDSS and annual rate of EDSS progression. Results:Compared with controls, MSnON eyes showed narrower central retinal venules (P = 0.014), increased arteriole-to-venule ratio (AVR) (P = 0.031), and reduced venular fractal dimension (FD) (P = 0.013). Disability correlated with increased venular caliber (P < 0.001), vessel density (P = 0.026), superficial vascular complex (SVC) capillary density (P < 0.001), and choroidal thickness (P = 0.006), and decreased AVR (P = 0.006) and choroidal vascularity index (P = 0.030). Annual EDSS progression was associated with increased arteriolar caliber (P = 0.001), AVR (P= 0.025), SVC capillary density (P < 0.001), foveal avascular zone (FAZ) volume (P < 0.001), and deep FAZ area (P < 0.001) and a decreased venular width gradient (P = 0.011) and FD (P < 0.001). Eyes with a history of ON eyes showed narrower venular caliber (P = 0.008), density (P = 0.012), and FD (P = 0.006). When MSON and MSnON are compared, ON affected only central arteriolar caliber (P = 0.010) and global SVC density (P = 0.010). Conclusions:Structural retinal and choroidal vascular alterations in MS were associated with disability and disease progression. These findings highlight the importance of retinal vascular assessments in the diagnosis, monitoring, and prognostication of MS, warranting confirmation in longitudinal studies. Financial Disclosures:Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Background: PREVENT-PE was a trial (n=8,094 women) comparing usual care with planned timed birth between 37 and 40 weeks’ gestation, based on risk stratification for preeclampsia (PE) at 35–36 weeks. In the intervention group, there was a 30% reduction in birth with PE, without an increase in emergency cesarean or neonatal unit (NNU) admission for ≥48 hours. Methods: This post hoc Bayesian analysis evaluates the impact of the intervention within each PE risk stratum in 4,734 participants from one study site where biobanked samples and hemodynamic assessments were available, permitting calculation of PE risk. Outcomes were birth with PE (primary trial outcome), and emergency cesarean and NNU for ≥48 hours (key secondary outcomes). Findings: Birth with PE incidence was 5.7% (95% credible interval [CrI], 4.7-6.8) in the control group and 4.1% [3.4-4.9] in the intervention group, for a posterior median incidence ratio of 0.72 [0.56-0.93] and a posterior probability-of-benefit of 0.995, consistent in direction and magnitude with those of the primary PREVENT-PE trial. However, overall benefit was driven primarily by the 39- and 40-week strata (posterior probabilities-of-benefit 0.985 and 0.992, respectively), with little evidence of excess NNU admission for ≥48 hours (posterior probability-of-harm 0.651 and 0.679, respectively), but a potential increase in emergency cesarean in the 39-week but not 40-week groups (posterior probability-of-harm 0.857 and 0.353). There was little evidence of benefit in 37- and 38-week strata (posterior probabilities-of-benefit, 0.463 and 0.622, respectively), and some evidence of excess NNU admission for ≥48 hours (posterior probabilities-of-harm, 0.746 and 0.825, respectively), but substantial uncertainty due to small numbers. Interpretation: Risk stratification with timed birth was associated with evidence of benefit in 39- and 40-week strata, with little suggestion of neonatal harm. In 37- and 38-week strata, there was no evidence of benefit, but potential neonatal harm consistent with early-term birth.

Objective To determine whether antihypertensive treatment of blood pressure levels of <140/90 mm Hg can reduce the incidence of pre-eclampsia.Design Secondary analysis of data from prospective cohort studies.Setting Three prospective screening studies of women who attended routine hospital maternity visits at 11-13 weeks' gestation, 1 February 2010 to 31 December 2016. Participants were from seven secondary care institutions in England.Participants 54 422 pregnancies screened at 11-13 weeks' gestation for pre-eclampsia and with blood pressure values available, that resulted in a liveborn or stillborn infant at ≥24 weeks' gestation.Main outcome measures Incidence of pre-eclampsia (overall, and at preterm or term gestational ages), according to modelled blood pressure lowering.Results The study population was ethnically diverse (17.3% black participants, 7.8% from South or East Asia, and 2.6% self-identified with more than one ethnic group). The Fetal Medicine Foundation competing risks model was used to calculate the expected risk of pre-eclampsia, based on maternal characteristics, mean arterial pressure, uterine artery pulsatility index, and placental growth factor. The expected risk of pre-eclampsia was used to calculate the expected incidence of pre-eclampsia. Reducing diastolic blood pressure from >85 mm Hg to a target of 85 mm Hg would mean that 4.8% of women would be offered antihypertensive drugs, with a potential relative risk reduction of 21.4% (absolute reduction of 2.9%) in any pre-eclampsia and 28.3% (absolute reduction of 1.4%) reduction in preterm pre-eclampsia. By reducing diastolic blood pressure from >80 mm Hg to a target of 80 mm Hg, 13.2% of women would receive antihypertensive drugs, with a potential relative risk reduction of 26.0% (absolute reduction of 2.3%) in any pre-eclampsia and 33.8% (absolute reduction of 1.0%) reduction in preterm pre-eclampsia. By reducing diastolic blood pressure from >75 mm Hg to a target of 75 mm Hg, 29.5% of women would receive antihypertensive drugs, with a potential relative risk reduction of 32.8% (absolute reduction of 2.1%) in any pre-eclampsia and 41.6% (absolute reduction of 0.8%) in preterm pre-eclampsia.Conclusions Lowering blood pressure from early pregnancy may reduce preterm and term pre-eclampsia. This finding requires evaluation in a definitive randomised trial.

Explainable AI (XAI) is essential for trust and transparency in deep learning, especially in medical imaging. Existing local explanation methods provide per-instance insights but fail to show whether similar explanations hold across samples of the same class. This limits global interpretability and demands time-consuming manual review by clinicians to trust models in practice. We introduce the Consensus Alignment Score (CAS), a novel metric that quantifies consistency of explanations at the class level. We also present ConsensusXAI, an open-source, model- and method-agnostic framework that evaluates explanation agreement quantitatively (via CAS) and qualitatively (through consensus heatmaps) per class. Unlike prior benchmarks, ConsensusXAI uses a latent-space clustering approach, Latent Consensus, to identify dominant explanation patterns, exposing biases and inconsistencies towards certain classes. Evaluated across two different medical imaging modalities for both correct and incorrect predictions on two different backbones, our method consistently reveals meaningful class-level insights, outperforming traditional consensus method i.e. SSIM, and enabling faster, more confident clinical adoption of AI models. The source code is available at https://github.com/a-haider1992/cas_toolbox.

Current brain imaging to detect silent brain infarctions (SBIs) is not feasible for the general population. Here, to overcome this challenge, we developed a retinal image-based deep learning system, DeepRETStroke, to detect SBI and refine stroke risk. We use 895,640 retinal photographs to pretrain the DeepRETStroke system, which encodes a domain-specific foundation model for representing eye-brain connections. Then, we validated the downstream clinical tasks of DeepRETStroke using 213,762 retinal photographs from diverse datasets across China, Singapore, Malaysia, the USA, the UK and Denmark to detect SBI and predict stroke events. DeepRETStroke performed well in internal validation datasets, with areas under the curve of 0.901 for predicting incident stroke and 0.769 for predicting recurrent stroke. External validations demonstrated consistent performances across diverse datasets. Finally, in a prospective study comprising 218 participants with stroke, we assessed the performance of DeepRETStroke compared with clinical traits in guiding strategies for stroke recurrence prevention. Altogether, the retinal image-based deep learning system, DeepRETStroke, is superior to clinical traits in predicting stroke events, especially by incorporating the detection of SBI, without the need for brain imaging.

BACKGROUND:In high-risk pregnancies, there is no reliable intervention to reduce term pre-eclampsia. We aimed to investigate the effect of screening for pre-eclampsia risk at 36 weeks' gestation and offering risk-stratified, planned, early-term birth. METHODS:PREVENT-PE was an open-label, adaptive (planned, for sample size), randomised controlled trial, done at two maternity hospitals in the UK. We included women (aged ≥16 years) with a singleton pregnancy, live fetus without major anomalies, and ability to provide informed consent, without pre-eclampsia or participation in conflicting trials. Consenting women were randomly assigned (by a central computerised service, 1:1, in random permuted blocks of variable size, stratified by site) to the intervention group (pre-eclampsia risk assessment and, for women with a pre-eclampsia risk ≥1 in 50, risk-stratified planned early-term birth) or control group (usual care at term). The primary outcome was birth with pre-eclampsia (International Society for the Study of Hypertension in Pregnancy criteria). This trial is registered with ISRCTN, ISRCTN41632964. FINDINGS:Of 11 280 women presenting for routine fetal ultrasound at 35+0 to 36+6 weeks' gestation between May 9, 2023, and June 7, 2024, 10 803 (95·8%) were eligible. Of 8136 women (75·3%) randomly assigned, six (0·1%) withdrew consent and 36 (0·4%) were randomly assigned in error, leaving 8094 (99·5%) in the final analyses: 4037 (49·9%) women in the intervention group and 4057 (50·1%) in the control group. 2098 (25·9%) of 8094 women self-reported non-White ethnicity and 5996 (74·1%) self-reported White ethnicity. Pre-eclampsia occurred in 158 (3·9%) of 4037 births in the intervention group and in 226 (5·6%) of 4057 in the control group (adjusted risk ratio 0·70 [95% CI 0·58-0·86]; intention-to-treat analysis with imputation). Serious adverse events did not differ between the intervention group (five [0·1%] of 4031) and control group (ten [0·2%] 4048; Fisher's exact test p=0·30). INTERPRETATION:Planned early-term birth based on risk stratification for pre-eclampsia reduced the incidence of pre-eclampsia, without increasing emergency caesarean section or neonatal care unit admission. FUNDING:Fetal Medicine Foundation.

Background/aims Within the UK, there are approximately 340 000 people who are registered as sight impaired (SI) or severely Sight Impaired (SSI), mainly due to age-related macular degeneration (AMD), diabetic eye disease (DED) and glaucoma. This study aimed to explore the association between certification of visual impairment (CVI) and socioeconomic deprivation.Methods Data from all CVI forms across Northern Ireland (NI) between 2018 to 2022 were used for analysis. Data collected included age, sex, visual acuity (logMAR), postcode and certification category. Deprivation measure was obtained using the Northern Ireland Multiple Deprivation Measure 2017 (NIMDM17). Patients were allocated a quintile from 1 (most deprived) to 5 (least deprived).Results Of the 1863 patients with VI who met the inclusion criteria, 1798 (97%) had postal codes recorded and therefore were allocated an NIMDM17 score. There were 755 patients in total, which were grouped into the most deprived and least deprived areas (357 and 398, respectively). Results showed that patients living in more deprived areas were significantly more likely to be certified as SI/SSI at a younger age than those living in less deprived areas (80.97 vs 85.77, respectively, p<0.001). This was seen in patients with AMD (85.1 vs 87.5, p=0.005) and DED (63.9 vs 69.8, p=0.013) but not in glaucoma (80.9 vs 84.1, p=0.073).Conclusion This study has shown that patients living in more deprived areas are more likely to be certified as SI or SSI at a significantly younger age compared with patients from less deprived areas across the certification database.

BACKGROUND/OBJECTIVES:This systematic literature review examines the quality of early clinical evaluation of artificial intelligence (AI) decision support systems (DSS) reported in glaucoma care. Artificial Intelligence applications within glaucoma care are increasing within the literature. For such DSS, there needs to be standardised reporting to enable faster clinical adaptation. In May 2022, a checklist to facilitate reporting of early AI studies (DECIDE-AI) was published and adopted by the EQUATOR network. METHODS:The Cochrane Library, Embase, Ovid MEDLINE, PubMed, SCOPUS, and Web of Science Core Collection were searched for studies published between January 2020 and May 2023 that reported clinical evaluation of DSS for the diagnosis of glaucoma or for identifying the progression of glaucoma driven by AI. PRISMA guidelines were followed (PROSPERO registration: CRD42023431343). Study details were extracted and were reviewed against the DECIDE-AI checklist. The AI-Specific Score, Generic-Item Score, and DECIDE-AI Score were generated. RESULTS:A total of 1,552 records were screened, with 19 studies included within the review. All studies discussed an early clinical evaluation of AI use within glaucoma care, as defined by the a priori study protocol. Overall, the DECIDE-AI adherence score was low, with authors under reporting the AI specific items (30.3%), whilst adhering well to the generic reporting items (84.7%). CONCLUSION:Overall, reporting of important aspects of AI studies was suboptimal. Encouraging editors and authors to incorporate the checklist will enhance standardised reporting, bolstering the evidence base for integrating AI DSS into glaucoma care workflows, thus help improving patient care and outcomes.

BACKGROUND:Gestational diabetes mellitus is a pregnancy complication that can be associated with increased risks of adverse maternal and neonatal outcomes. Optimal glycemic control remains challenging for many patients despite the existing management strategies. Ursodeoxycholic acid is commonly used for cholestasis of pregnancy and has shown potential metabolic benefits, including improved insulin sensitivity and reduced inflammation. We hypothesize that ursodeoxycholic acid may improve glycemic control in gestational diabetes mellitus. OBJECTIVE:This study aimed to compare treatment with ursodeoxycholic acid vs placebo for improving maternal glycemia in gestational diabetes mellitus. STUDY DESIGN:This was a single-site, randomized, double-blind, placebo-controlled trial of ursodeoxycholic acid in 113 women with gestational diabetes mellitus at 24 to 28 weeks' gestation. The primary outcome was maternal fasting blood glucose concentration at 35+0 to 37+6 weeks' gestation. RESULTS:The primary outcome did not differ significantly between groups when evaluated by intention to treat analysis (treatment effect, 0.98; 95% confidence interval, 0.92-1.05; P=.61). There were no differences in maternal or fetal secondary outcomes, including maternal weight change, need for insulin treatment, birthweight centile, proportion of large or small for gestational age infants, neonatal hypoglycemia, or admission to the neonatal unit. A prespecified secondary analysis measured serum concentrations of ursodeoxycholic acid using ultra-performance liquid chromatography-tandem mass spectrometry and showed that participants taking larger numbers of tablets had higher serum concentrations of ursodeoxycholic acid. Post hoc analysis revealed no difference in the rate of fasting blood glucose concentrations at or above the recommended target of 90 mg/dL according to intention to treat analysis (5/50 [10.0%] vs 8/53 [15.1%]; risk ratio, 0.66; 95% confidence interval, 0.23-1.89; P=.557). However, among patients with serum ursodeoxycholic acid ≥0.5 µmol/L (indicating adherence), fewer patients had fasting glucose levels above the target (2/42 [4.8%] vs 11/57 [19.3%]; risk ratio, 0.25; 95% confidence interval, 0.06-1.06; P=.039). CONCLUSION:This trial demonstrated no difference in fasting glycemia between women with gestational diabetes mellitus treated with ursodeoxycholic acid and those treated with placebo. However, those with elevated serum ursodeoxycholic acid concentrations were more likely to have fasting blood glucose concentrations below the recommended thresholds, suggesting potential benefit of further investigation.

Background/aims To investigate the landscape to support Europe-wide collaborative real-world data (RWD) collection, exploring whether required resources are available to glaucoma clinicians.Methods This cross-sectional study employed a two-phase method, including two consecutive electronic questionnaires. For phase I, a survey was distributed to all European Glaucoma Society members inquiring about use of electronic medical records (EMRs). For phase II, EMR users who expressed their interest in contributing to a European RWD were surveyed regarding EMR data characteristics and technical resources.Results Phase I garnered responses from 201 glaucoma specialists, with 68% reporting use of an EMR. Among them, 47% and 65% reported over 90% of coding rates for diagnoses and medications, respectively. Despite the EMR making available report copies for perimetry (27%) and optical coherence tomography (OCT) (37%), automatic data extraction capabilities were limited to only 10% and 8%. 47 participants responded to phase II. Their practices saw an average of 415 glaucoma patients per month. Date of intraocular pressure (IOP) measurement was the most commonly available structured data (67%), followed by visual acuity (59%) and IOP (56%). 46% had access to central computing systems, with 41% reporting past experience in federated learning initiatives and 73% willing contribution. 34% reported involvement in the development of artificial intelligence algorithms.Conclusion This study highlights a noteworthy EMR use among European glaucoma specialists, with structured data for metrics like IOP, but identifies gaps in data extraction capabilities. Major challenges include EMR heterogeneity, lack of standardised outcomes, healthcare system differences and cost. A combined approach using structured coding, digitised EMR data and semistructured OCT/perimetry data is recommended.