BMJ Open Ophthalmology最新文献

Objective: To assess the safety and tolerability of avacincaptad pegol (ACP), a Food and Drug Administration-approved therapy for geographic atrophy, administered in combination with ranibizumab, an approved therapy for neovascular age-related macular degeneration (nAMD), in patients with nAMD.

Methods: The phase 1 study (NCT00709527) was a two-part, ascending-dose and parallel-group, open-label trial that assessed the safety, tolerability and pharmacokinetic profile of monthly intravitreal injections of ACP (0.03, 0.3, 1, 2, 3 mg) in combination with ranibizumab (0.5 mg) on the same day in treatment-naïve patients with nAMD (n=43 patients received a maximum of 6 injections). The phase 2a study (NCT03362190) was an open-label trial assessing the 6-month safety of intravitreal injections of ACP administered in combination with ranibizumab in treatment-naïve patients with nAMD (n=64). Patients received either ACP 2 mg or 4 mg either 14 days or 1 month apart, given on the same day or 2 days after ranibizumab. Primary outcome measures were safety and tolerability.

Results: During the phase 1 study, there were no dose-limiting toxicities at any dose level. In both studies, ocular treatment-emergent adverse events were mostly mild or moderate, with the most reported events related to the injection procedure. There were no clinically significant increases in intraocular pressure or cumulative increases with multiple injections over time. There were no safety issues identified through measurement of visual acuity.

Conclusions: Coadministration of ACP and ranibizumab in treatment-naïve patients with nAMD was well tolerated across different dosing regimens with no new safety issues based on results from two independent studies.

Objective: To summarise global peer-reviewed primary research on eye health published from 2000 to 2019.

Methods and analysis: We used the 'explode eye disease' function on MEDLINE to obtain all articles reporting primary research studies on eye health published between 1 January 2000 and 31 December 2019. We were intentionally broad and included population, clinical, animal and laboratory studies. We categorised the main eye condition of the paper from Medical Subject Headings (MeSH) terms, and the country of the study from the first country listed in the abstract (or if this was absent, the affiliation of the first author). A validated algorithm was used to assign gender to authors.

Results: We included 158 697 publications from 178 countries. Across the period, annual research output increased globally (4.2% per annum, 5057 publications in 2000 to 10 875 in 2019) and in 20 of 21 regions. There was substantial geographical maldistribution, with research output ranging from 1.0 publication/million population in Central Sub-Saharan Africa to 165.8/million in Australasia; 70% of research identified was conducted in high-income countries (n=1 11 417). 42% of publications focused on one of the five leading causes of vision impairment. Of the 789 463 authorships assigned a gender, women held 33% of all (n=261 636/789 463), 36% of first (n=47 729/131 664) and 24% of last authorships (n=31 720/129 800). Women formed 50% of authorship teams when the last author was a woman (IQR 38-71%), compared with 20% of teams when the last author was a man (IQR 0-40%).

Conclusion: The annual research output doubled globally over the two decades, with a disproportionate output from high-income countries and slow progress towards gender parity. The main limitations of our study included the use of a single database, which may have led to an underestimation of all outputs, particularly from low- or middle-income countries.

Objective: Metformin has been identified as a potential treatment for age-related macular degeneration (AMD). Photographic screening for diabetic retinopathy provides an opportunity to conduct a case-control study with systematic AMD grading. We aimed to investigate associations between metformin use and incidence and progression of AMD at different grades.

Methods and analysis: We randomly sampled 2600 participants from 10 336 people aged ≥50 years with diabetes who attended retinopathy screening in 2011 (baseline) and were enrolled to the Individualised Screening for Diabetic Retinopathy study. 2545 of these participants had type 2 diabetes and gradable fundus photographs at baseline, which were graded using modified age related eye disease study grading. We used data including those on metformin prescription from general practitioner records. We used multivariate logistic regression to investigate associations between metformin and incidence or progression of early, intermediate and late AMD.

Results: Of 2545 participants, 2089 attended and had gradable fundus images on year 5. Metformin was associated with reduced incidence of intermediate AMD by 5 years after adjusting for confounders (complete record OR 0.63, 95% CI 0.43 to 0.92, p=0.02). In univariate analysis, metformin was associated with reduced incidence of late AMD (OR 0.43, 95% CI 0.21 to 0.88, p=0.02) but this did not remain significant after adjusting for age and sex. The numbers progressing to late AMD were small. There was no association between metformin and the incidence of early AMD.

Conclusion: We have found a significant association between metformin use and reduction in incidence of intermediate AMD by 37% in people with diabetes over 5 years. Previous epidemiological studies of metformin and AMD have used secondary data on AMD. In this observational study, there were baseline differences between groups, although significant findings remained after adjusting for important confounders. Given metformin's anti-ageing therapeutic effects, the reduction in risk is plausible and warrants prospective clinical trials.

Aims: Evaluate the efficacy and safety of faricimab for treatment-naïve and pre-treated diabetic macular oedema (DME) in a real-world setting.

Methods: This multicentre, retrospective cohort study examined consecutive DME patients treated with faricimab for ≥1 year. Data were collected at predefined time points. Primary outcomes were mean changes in corrected visual acuity (VA), centre-point retinal thickness (CRT) and central subfield thickness (CST) and treatment intervals and adverse events (AEs).

Results: 184 eyes with DME were included: 61 (33.2%) were treatment-naïve, and 123 (66.8%) were pretreated. In treatment-naïve eyes, VA improved from 69.7±15 Early Treatment of Diabetic Retinopathy Study letters at baseline to 73.9±14.1 after 12 months (p=0.014), while it remained stable in pretreated eyes (71.2±14.2 vs 73.0±12.9; p=0.14). CST decreased significantly in both groups (treatment-naïve (366.1±108.3 µm to 316.4±113.5 µm, p<0.001); pretreated (339.1±93.1 µm to 298.3±65.8 µm, p<0.001)). Thirty-one percent of naïve eyes and 21.1% of pretreated eyes were completely dry after 12 months. In treatment-naïve eyes, the mean treatment interval was 12.7±6.4 weeks at 12 months. In pretreated eyes, the interval increased from 6.0±3.0 to 7.8±3.6 weeks (p<0.001). Over 12 months, 8.1±2.1 and 9.4±2.5 injections were administered to naïve and pretreated eyes, respectively (p<0.001). Of the five recorded AEs, two cases of non-infectious intraocular inflammation and one cerebrovascular event were reported.

Conclusion: Over 12 months, faricimab demonstrated good efficacy and safety in both treatment-naïve and pretreated eyes with DME. There was a reduction in CST and improved VA in treatment-naïve eyes and stable VA in pretreated eyes. The low number of AEs supports a favourable risk-benefit profile of faricimab in a real-world setting.

Objective: To analyse the data of a cohort of patients with conjunctival melanoma in order to analyse risk and guarding factors and to investigate the impact on metastatic disease with and without adjuvant therapy.

Methods and analysis: We have retrospectively analysed the impact of clinical aspects and adjuvant therapies after tumour excision or biopsy in 167 patients cared for at the University Hospital Essen and the University Hospital Tübingen, Germany. Clinical as well as histopathological data and therapeutic approaches were analysed with regard to regional (lymphatic) and/or distant (haematogenous) spread during follow-up. The Kaplan-Meier estimate method was used to analyse survival and the Cox regression hazard model to define the probability of metastases depending on different factors. P value of <0.05 was considered statistically significant.

Results: 167 cases of malignant conjunctival melanoma were retrospectively analysed. The patients received treatment and were followed up for 78.3±67.5 months. Local tumour recurrence occurred in n=79 patients (47.3%) after a mean of 41.5±70.33 months. 30 patients (37.9%) with a recurrence had not received adjuvant therapy. The overall rate of metastasis was 24.5% (n=41). In n=31 cases, regional lymphatic metastases were found after a follow-up of 48.9±63.5 months; in n=24 cases, distant metastases were found, occurring after 55.5±67.4 months. In n=14, the metastatic disease took both pathways. Ruthenium-106 brachytherapy performed in localised melanoma of the bulbar conjunctiva showed a relevant effect of decreasing the risk for haematogenous metastases by 74% (HR=0.256, p=0.003). 4 out of 54 patients developed distant metastases.

Conclusion: In conjunctival melanoma, it is important to perform an adjuvant therapy after excision. This reduces not only local recurrences but also significantly the risk for haematopoietic spread.

Objectives: To evaluate the performance of reasoning large language models (LLMs) with human-like thinking in ophthalmic question answering.

Methods: We evaluated two state-of-the-art open-source reasoning LLMs (DeepSeek-R1 and QwQ-32B) and one conventional non-reasoning LLM (LLaMA-3.3-70B-Instruct) models on ophthalmology questions, assessing not only answer accuracy (ACC) but also the quality of their reasoning processes. First, we curated MedQA-Eye, a dataset of 967 ophthalmology questions across 10 subspecialties, 3 scenarios, 5 medical entities and 3 languages. Second, we proposed a novel framework considering human thinking patterns essential to medical practice to evaluate the thinking performance of reasoning LLMs on MedQA-Eye.

Results: DeepSeek-R1 demonstrated superior overall ACC (90.59%, 95% CI 88.59% to 92.27%) to LLaMA-3.3-70B-Instruct (87.90%, 95% CI 85.69% to 89.81%, p=0.015) and QwQ-32B (84.28%, 95% CI 81.85% to 86.44%, p<0.001) with performance varying across subspecialties. Analysis of reasoning LLMs revealed incorrect logical inference as the primary point of failure, accounting for 93.41%-94.74% of incorrectly answered questions. We further quantified semantic uncertainty in reasoning LLM thinking as a predictor of answer reliability. DeepSeek-R1 exhibited lower semantic uncertainty (1.04±3.63) compared with QwQ-32B (4.31±40.70), p<0.001.

Conclusion: Reasoning LLMs demonstrated superior performance in ophthalmology question answering, with DeepSeek-R1 achieving the highest ACC. Our findings demonstrate that reasoning LLM can better simulate human-like thinking processes compared with conventional non-reasoning LLM, suggesting its potential for more trustworthy LLM systems in ophthalmology.

Objective: Retinal biomarkers accessible via non-invasive optical coherence tomography (OCT) could facilitate early detection of Alzheimer's disease (AD), complementing current invasive or costly diagnostic methods. This review evaluates the evidence for spectral-domain OCT (SD-OCT) and OCT angiography (OCT-A) in identifying retinal changes associated with preclinical and early AD.

Methods and analysis: We conducted a systematic review registered in PROSPERO and aligned with Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 guidelines. PubMed/MEDLINE was searched up to April 2025, complemented by reference list screening and citation tracking. Eligible studies assessed SD-OCT and/or OCT-A in biomarker-defined preclinical or early AD, mild cognitive impairment or mild AD. Data were synthesised narratively by disease stage, and methodological quality was appraised with the Newcastle-Ottawa Scale.

Results: 22 studies met inclusion criteria. Reported alterations included thinning of the peripapillary retinal nerve fibre layer and retinal ganglion cell layer, macular and choroidal thickness changes and microvascular alterations on OCT-A. However, findings were heterogeneous: some studies observed early thickening or increased vascular density, possibly reflecting inflammatory or compensatory mechanisms, while others reported thinning and rarefaction more consistent with neurodegeneration. Most studies were of moderate quality, limited by small sample sizes, cross-sectional designs and incomplete control for ocular/systemic confounders.

Conclusion: SD-OCT and OCT-A hold promise as candidate biomarkers of early AD, but current evidence remains variable, non-specific and methodologically constrained. Further research is needed to standardise imaging protocols, validate findings in biomarker-confirmed longitudinal cohorts and compare OCT-based measures across dementia subtypes. Integration with other biomarkers (eg, plasma or metabolomics) may improve diagnostic specificity and support translation of OCT/OCT-A into clinical practice.

Prospero registration number: CRD42024600456.

Aim: To identify combinations of up to three visual function tests with the best performance for classifying diabetic retinopathy (DR) severity stage. To describe in detail the measurements from a comprehensive set of visual function tests.

Methods: 1901 eyes (1032 participants) underwent nine visual function tests. Fundus, ultra-widefield and optical coherence tomography images were graded for DR and diabetic macular oedema (DMO). Three classification tasks were set: (1) distinguishing diabetes mellitus (DM) no DR from healthy with no DM, (2) DR no DMO from DM no DR and (3) DR with DMO from DR no DMO. Ensemble machine learning models for all one-way, two-way and three-way combinations of visual function variables were compared using area under the curve (AUC).

Results: The top 30 models for each task achieved high accuracy, with AUC ≥0.94. For task 1, 17/30 top models contained distance visual acuity. Pelli-Robson contrast sensitivity and low luminance visual acuity also featured highly. For task 2, 19/30 models contained mesopic microperimetry. Near visual acuity, matrix microperimetry and reading index featured highly. For task 3, 17/30 models contained distance visual acuity. Smith-Kettlewell low luminance near visual acuity and near visual acuity featured highly. In a subset of eyes where perimetry was not performed, reading index featured in 22, 21 and 22 of the top models for tasks 1, 2 and 3 respectively.

Conclusions: These findings will enable researchers and those planning clinical trials to select the best combination of visual function tests for distinguishing stages of diabetic eye disease.

Introduction: Ocular diseases, though common in intensive care unit (ICU) patients, are often overlooked due to prioritisation of organ failure management. Their comorbidities, clinical features and prognostic effects require large-scale investigation.

Measurements and main results: This study used data from the Medical Information Mart for Intensive Care IV database (V. 2.2). Random forest models were used to determine the importance of factors for 1-year mortality and in-hospital mortality. Multivariate logistic regression models were employed to identify the independent risk factors associated with in-hospital mortality in acute ocular disease subgroups. A total of 40 149 patients were identified in this study as eligible for the final analysis. The median age was 67.1 years, 22 734 (56.6%) were male and 3920 (9.8%) had ocular disease. 401 had acute ocular diseases, 2563 had chronic ocular diseases, 948 had neurological ocular diseases and 26 suffered from traumatic ocular diseases. Patients with acute ocular diseases had significantly higher proportions of acute kidney injury (76.8%, p=0.024), intensive mechanical ventilation (51.9%, p=0.015), norepinephrine (23.4%, p=0.008), neuro block (5.0%, p=0.001), sedation (59.6%, p=0.004) and sepsis (59.6%, p<0.001) than those without. There were different trends among cohorts with or without chronic ocular disease. Patients with chronic ocular diseases had a significantly higher 1-year mortality after ICU admission than those without (log-rank p=0.0001). Acute non-traumatic haemorrhagic ocular disease was independently associated with increased in-hospital mortality (OR 2.69, 95% CI 1.24 to 5.84) among ICU patients with acute ocular complications.

Conclusions: ICU patients with chronic ocular diseases exhibited higher long-term mortality. Acute ocular diseases may indicate a higher severity of illness in critically ill patients. Within the subgroup of acute ocular diseases, acute non-traumatic haemorrhagic ocular diseases were independently associated with increased in-hospital mortality. Limitations include retrospective under-ascertainment of asymptomatic cases and inability to distinguish aetiologies.

Objectives: Retinopathy of prematurity (ROP) is a leading cause of blindness in children worldwide, requiring more efficient models to help predict treatment-requiring ROP. Our study aimed to develop a new prediction model for ROP occurrence and severity, named NLP-ROPCare, using natural language processing (NLP).

Methods and analysis: A retrospective observational study. Infants with a gestational age ≤32 weeks or birth weight ≤2000 g were collected in Guangdong Women and Children Hospital from 2013 to 2022, including 3922 preterm infants with 1106 patients with ROP. Four pretrained language models - BERT (Bidirectional Encoder Representations from Transformers), RoBERTa (Robustly Optimized BERT pretraining Approach), MC-BERT (language pre-training via a Meta Controller) and NEZHA (NEural contextualiZed representation for CHinese lAnguage understanding) - were used for development of NLP prediction models based on free-form texts in the admission notes. For comparison, two machine learning methods (Random Forest and Support Vector Machine) were used to construct prediction models based on 20 structured characteristics previously extracted from the admission notes. Performance evaluating metrics included accuracy, precision, recall, F1 score and area under the curve (AUC).

Results: The NLP prediction models for ROP occurrence outperformed those for severity. The NEZHA model demonstrated the highest accuracy in predicting ROP occurrence, achieving an F1 score of 89.35% and an AUC of 0.90. Its performance was also better than two machine learning models whose highest F1 was 78% with an AUC equal to 0.87. In addition, the F1 score of RoBERTa (78.44%) was slightly higher than that of NEZHA (77.81%) for predicting ROP severity, and the AUC of RoBERTa also achieved the highest 0.91.

Conclusion: The NLP-ROPCare combines language models NEZHA and RoBERTa to enable early prediction of ROP occurrence and severity based on unstructured free-form texts in the admission notes of preterm infants, highlighting its value in early prevention of ROP. Further external validation should be carried out to better adjust the model.