Vision of the future: large language models in ophthalmology.

PURPOSE OF REVIEW: Large language models (LLMs) are rapidly entering the landscape of medicine in areas from patient interaction to clinical decision-making. This review discusses the evolving role of LLMs in ophthalmology, focusing on their current applications and future potential in enhancing ophthalmic care. RECENT FINDINGS: LLMs in ophthalmology have demonstrated potential in improving patient communication and aiding preliminary diagnostics because of their ability to process complex language and generate human-like domain-specific interactions. However, some studies have shown potential for harm and there have been no prospective real-world studies evaluating the safety and efficacy of LLMs in practice. SUMMARY: While current applications are largely theoretical and require rigorous safety testing before implementation, LLMs exhibit promise in augmenting patient care quality and efficiency. Challenges such as data privacy and user acceptance must be overcome before LLMs can be fully integrated into clinical practice.

Appropriateness of Ophthalmology Recommendations From an Online Chat-Based Artificial Intelligence Model.

Objective: To determine the appropriateness of ophthalmology recommendations from an online chat-based artificial intelligence model to ophthalmology questions. Patients and Methods: Cross-sectional qualitative study from April 1, 2023, to April 30, 2023. A total of 192 questions were generated spanning all ophthalmic subspecialties. Each question was posed to a large language model (LLM) 3 times. The responses were graded by appropriate subspecialists as appropriate, inappropriate, or unreliable in 2 grading contexts. The first grading context was if the information was presented on a patient information site. The second was an LLM-generated draft response to patient queries sent by the electronic medical record (EMR). Appropriate was defined as accurate and specific enough to serve as a surrogate for physician-approved information. Main outcome measure was percentage of appropriate responses per subspecialty. Results: For patient information site-related questions, the LLM provided an overall average of 79% appropriate responses. Variable rates of average appropriateness were observed across ophthalmic subspecialties for patient information site information ranging from 56% to 100%: cataract or refractive (92%), cornea (56%), glaucoma (72%), neuro-ophthalmology (67%), oculoplastic or orbital surgery (80%), ocular oncology (100%), pediatrics (89%), vitreoretinal diseases (86%), and uveitis (65%). For draft responses to patient questions via EMR, the LLM provided an overall average of 74% appropriate responses and varied by subspecialty: cataract or refractive (85%), cornea (54%), glaucoma (77%), neuro-ophthalmology (63%), oculoplastic or orbital surgery (62%), ocular oncology (90%), pediatrics (94%), vitreoretinal diseases (88%), and uveitis (55%). Stratifying grades across health information categories (disease and condition, risk and prevention, surgery-related, and treatment and management) showed notable but insignificant variations, with disease and condition often rated highest (72% and 69%) for appropriateness and surgery-related (55% and 51%) lowest, in both contexts. Conclusion: This LLM reported mostly appropriate responses across multiple ophthalmology subspecialties in the context of both patient information sites and EMR-related responses to patient questions. Current LLM offerings require optimization and improvement before widespread clinical use.

A Comparative Study of Large Language Models, Human Experts, and Expert-Edited Large Language Models to Neuro-Ophthalmology Questions.

BACKGROUND: While large language models (LLMs) are increasingly used in medicine, their effectiveness compared with human experts remains unclear. This study evaluates the quality and empathy of Expert + AI, human experts, and LLM responses in neuro-ophthalmology. METHODS: This randomized, masked, multicenter cross-sectional study was conducted from June to July 2023. We randomly assigned 21 neuro-ophthalmology questions to 13 experts. Each expert provided an answer and then edited a ChatGPT-4-generated response, timing both tasks. In addition, 5 LLMs (ChatGPT-3.5, ChatGPT-4, Claude 2, Bing, Bard) generated responses. Anonymized and randomized responses from Expert + AI, human experts, and LLMs were evaluated by the remaining 12 experts. The main outcome was the mean score for quality and empathy, rated on a 1-5 scale. RESULTS: Significant differences existed between response types for both quality and empathy (P < 0.0001, P < 0.0001). For quality, Expert + AI (4.16 ± 0.81) performed the best, followed by GPT-4 (4.04 ± 0.92), GPT-3.5 (3.99 ± 0.87), Claude (3.6 ± 1.09), Expert (3.56 ± 1.01), Bard (3.5 ± 1.15), and Bing (3.04 ± 1.12). For empathy, Expert + AI (3.63 ± 0.87) had the highest score, followed by GPT-4 (3.6 ± 0.88), Bard (3.54 ± 0.89), GPT-3.5 (3.5 ± 0.83), Bing (3.27 ± 1.03), Expert (3.26 ± 1.08), and Claude (3.11 ± 0.78). For quality (P < 0.0001) and empathy (P = 0.002), Expert + AI performed better than Expert. Time taken for expert-created and expert-edited LLM responses was similar (P = 0.75). CONCLUSIONS: Expert-edited LLM responses had the highest expert-determined ratings of quality and empathy warranting further exploration of their potential benefits in clinical settings.

A Comparative Study of Responses to Retina Questions from Either Experts, Expert-Edited Large Language Models, or Expert-Edited Large Language Models Alone.

Objective: To assess the quality, empathy, and safety of expert edited large language model (LLM), human expert created, and LLM responses to common retina patient questions. Design: Randomized, masked multicenter study. Participants: Twenty-one common retina patient questions were randomly assigned among 13 retina specialists. Methods: Each expert created a response (Expert) and then edited a LLM (ChatGPT-4)-generated response to that question (Expert + artificial intelligence [AI]), timing themselves for both tasks. Five LLMs (ChatGPT-3.5, ChatGPT-4, Claude 2, Bing, and Bard) also generated responses to each question. The original question along with anonymized and randomized Expert + AI, Expert, and LLM responses were evaluated by the other experts who did not write an expert response to the question. Evaluators judged quality and empathy (very poor, poor, acceptable, good, or very good) along with safety metrics (incorrect information, likelihood to cause harm, extent of harm, and missing content). Main Outcome: Mean quality and empathy score, proportion of responses with incorrect information, likelihood to cause harm, extent of harm, and missing content for each response type. Results: There were 4008 total grades collected (2608 for quality and empathy; 1400 for safety metrics), with significant differences in both quality and empathy (P < 0.001, P < 0.001) between LLM, Expert and Expert + AI groups. For quality, Expert + AI (3.86 ± 0.85) performed the best overall while GPT-3.5 (3.75 ± 0.79) was the top performing LLM. For empathy, GPT-3.5 (3.75 ± 0.69) had the highest mean score followed by Expert + AI (3.73 ± 0.63). By mean score, Expert placed 4 out of 7 for quality and 6 out of 7 for empathy. For both quality (P < 0.001) and empathy (P < 0.001), expert-edited LLM responses performed better than expert-created responses. There were time savings for an expert-edited LLM response versus expert-created response (P = 0.02). ChatGPT-4 performed similar to Expert for inappropriate content (P = 0.35), missing content (P = 0.001), extent of possible harm (P = 0.356), and likelihood of possible harm (P = 0.129). Conclusions: In this randomized, masked, multicenter study, LLM responses were comparable with experts in terms of quality, empathy, and safety metrics, warranting further exploration of their potential benefits in clinical settings. Financial Disclosures: Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of the article.

Determination of the Trans-Lamina Cribrosa Pressure Difference in a Community-Based Population and its Association with Open-Angle Glaucoma.

PURPOSE: To determine the trans-lamina cribrosa pressure difference (TLCPD) in a cohort of normal community-based patients and the relationship to primary open-angle glaucoma (POAG) and normal-tension glaucoma (NTG). DESIGN: Retrospective cohort study of the Mayo Clinic Study of Aging. PARTICIPANTS: The Mayo Clinic Study of Aging is a prospective study evaluating the normal aging population. METHODS: Mayo Clinic Study of Aging patients who underwent routine lumbar puncture (LP) studies with eye examinations were reviewed. The trans-lamina cribrosa pressure difference was calculated in 2 contexts of intraocular pressure (IOP): (1) maximum IOP at eye visit closest in time to the LP (closest-in-time TLCPD); and (2) IOP before IOP-lowering treatment (pretreatment IOP and pretreatment TLCPD) in POAG and NTG patients. Glaucoma patients without POAG or NTG were excluded. Regression analyses were performed to determine the relationship with glaucoma. MAIN OUTCOME MEASURES: IOP, intracranial pressure, TLCPD, POAG, normal-tension glaucoma (NTG) diagnosis, glaucoma parameters. RESULTS: Five hundred forty-eight patients were analyzed. Of these, there were 38 treated glaucoma patients (14 POAG and 24 NTG) and 510 nonglaucomatous patients. Cerebral spinal fluid (CSF) opening pressure was 155.0 ± 42.2 mmH2O in nonglaucomatous patients, 144.0 ± 34.0 mmH2O in POAG (P = 0.15 vs. nonglaucomatous patients), and 136.6 ± 29.3 mmH2O in NTG (P = 0.017 vs. nonglaucomatous patients). Intraocular pressure was 15.47 ± 2.9 mmHg in nonglaucomatous patients, 26.6 ± 3.7 mmHg in POAG, and 17.4 ± 3.4 mmHg in NTG. The closest-in-time TLCPD in the nonglaucomatous cohort was 4.07 ± 4.22 mmHg, which was lower than both the POAG cohort (7.19 ± 3.6 mmHg) and the NTG cohort (5.79 ± 4.5 mmHg, P = 0.04). Pretreatment TLCPD for the overall glaucoma cohort was 10.57 ± 6.1 mmHg. The POAG cohort had a higher pretreatment TLCPD (16.05 ± 5.2 mmHg) than the NTG cohort (7.37 ± 4.1 mmHg; P < 0.0001). Closest-in-time TLCPD for the nonglaucoma cohort (4.07± 4.2 mmHg) was significantly lower than pretreatment TLCPDs for both POAG (16.05 ± 5.2 mmHg; P < 0.0001) and NTG (7.37 ± 4.1 mmHg; P < 0.0001) cohorts. CONCLUSIONS: This study establishes the baseline TLCPD in a large cohort of normal, community-based patients. The differences in regression analysis between TLCPD and IOP suggests NTG pathophysiology is partially driven by TLCPD, but is also likely multifactorial. FINANCIAL DISCLOSURE(S): Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Trends in Myopia and High Myopia from 1966 to 2019 in Olmsted County, Minnesota.

PURPOSE: To investigate the prevalence of myopia and high myopia in Olmsted County, Minnesota, from 1966 to 2019. DESIGN: Retrospective, cross-sectional, trend study. METHODS: Manifest refractions or lens prescription were converted to spherical equivalent (SE) to estimate the prevalence of adult myopia and high myopia. Age, sex, race, and visual acuity were recorded. Subjects with an SE of -0.5 diopters (D) or less were considered to have myopia; those with an SE of -6.0 D or less were considered to have high myopia. Exclusion criteria included visually significant cataract, pseudophakia, prior refractive surgery, or age less than 18 years. RESULTS: Among 81,706 sampled subjects, the myopia prevalence increased from 33.9% (95% CI, 31.1-36.8) in the 1960s to 57.1% (95% CI, 56.6-57.6) in the 2010s (P < .001). The high myopia prevalence increased from 2.8% (95% CI, 1.95-3.98) in the 1960s to 8.3% (95% CI, 8.08-8.62) in the 2010s (P < .001). Both male (32.0%-55.1%, P < .001) and female (40.6%-58.5%, P < .001) subjects experienced increasing myopia prevalence from the 1960s to the 2010s; both male (2.6%-7.4%, P < .001) and female (3.4%-9.1%, P < .001) subjects also had higher high myopia prevalence rates from the 1960s through the 2010s. Increasing myopia and high myopia prevalence was detected by decade in nearly all age groups (excluding 18- to 24-year-old high myopia subjects). White and Asian subjects had the highest myopia prevalence, whereas Black subjects had the lowest. From the 2000s to the 2010s, White (53.3%-57.0%, P < .001) and Black (41.0%-47.0%, P = .001) subjects had significant increases in myopia prevalence. The mean SE decreased from the 1960s (-0.42 D; 95% CI, -0.59 to +2.49) to the 2010s (-1.85 D; 95% CI, -1.88 to +2.96) (P < .001). CONCLUSIONS: From 1966 to 2019 in Olmsted County, Minnesota, there was a 68% and 199% increase in myopia and high myopia prevalence, respectively.

Trends in Ophthalmology Practice Consolidation: 2015-2022.

PURPOSE: To quantify trends in ophthalmology practice consolidation in the United States. DESIGN: A retrospective cross-sectional study. PARTICIPANTS: Providers in the Centers for Medicare and Medicaid Services (CMS) National Downloadable File with a primary specialty designation of ophthalmology. METHODS: We used the CMS database to determine national practice consolidation trends in ophthalmology on individual physician and group practice levels and analyzed by region, sex, and years spent in practice. We used the Cochran-Armitage test to determine the statistical significance of practice size differences between 2015 and 2022. MAIN OUTCOME MEASURES: Temporal practice size trends for physicians and practices in ophthalmology and regional, sex-specific, and age-related trends. RESULTS: Between 2015 and 2022, the number of ophthalmologists decreased from 17 656 to 17 615 (-0.2%), whereas the number of practices decreased from 7149 to 5890 (-18%). The percentage of ophthalmologists in practices of 1 to 2 members decreased from 35% to 28%, whereas those in groups of 50 or more increased from 7% to 11%. The percentage of practices with 1 to 2 members decreased from 75% to 71%, and those with 50 or more increased from 0.2% to 0.4%. Consolidation trends were significant on individual ophthalmologist (P < 0.001) and group practice (P < 0.001) levels. All regions, sexes, and subgroups of years spent in practice demonstrated consolidation (P < 0.001). The Northeast showed the greatest increase in groups of 50 or more physicians (+7%) between 2015 and 2022. Proportionally fewer female than male ophthalmologists were associated with practice sizes of 1 to 2 members in 2015 (29% and 36%, respectively) and 2022 (23% and 30%, respectively). Proportionally fewer ophthalmologists with 0 to 10 years of experience in practice were associated with practice sizes of 1 to 2 members than those with more than 30 years in practice in 2015 (18% and 48%, respectively) and 2022 (14% and 40%, respectively). CONCLUSIONS: Ophthalmology has undergone practice consolidation from 2015 to 2022. A decrease in the proportion of physicians affiliated with smaller practice sizes seems to have occurred. FINANCIAL DISCLOSURE(S): Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Blurred Vision After a Kidney Transplant.

A patient with cytomegalovirus viremia who underwent kidney transplant developed floaters in both eyes that did not improve after painful intravitreal injections of foscarnet and ganciclovir. What would you do next?

Seeing Blue Dots After COVID-19 Infection.

A 12-year-old female individual receives a diagnosis of SARS-CoV-2 and reports bilateral blurry vision, large blue paracentral scotomata, and a migraine without a scintillating scotoma. What would you do next?

Oleander-Associated Keratitis and Uveitis.

PURPOSE: Oleander is a poisonous plant with extensively documented systemic side effects; however, oleander's ophthalmic side effects have not been detailed in the literature. We report a case of oleander-associated keratitis with subsequent corneal edema and anterior uveitis. METHODS: This is a case report and review of relevant literature. RESULTS: A 58-year-old woman presented with large corneal epithelial defect after being struck in the eye with an oleander leaf. Despite treatment with topical moxifloxacin, she developed severe corneal edema and anterior uveitis. A diagnosis of oleander-associated ocular inflammation with secondary corneal edema was made, given the temporal relationship, and treatment was initiated with topical prednisolone and cyclopentolate. However, the corneal edema and inflammation continued to progress until oral prednisone and topical difluprednate were initiated. Visual acuity, anterior uveitis, and corneal edema significantly improved with aggressive immunomodulation. Follow-up at 1 month confirmed complete recovery of symptoms, corneal edema and anterior uveitis. CONCLUSIONS: Severe corneal edema and anterior uveitis can be associated with oleander exposure. Aggressive treatment with oral and topical steroids may be required without persistent sequelae at the 5-month follow-up. Ophthalmologists should consider this inflammatory reaction if patients experience ocular exposure to oleander.

Case report: acute myocarditis following the second dose of mRNA-1273 SARS-CoV-2 vaccine.

BACKGROUND: The SARS-CoV-2 pandemic has led to the development of the first mRNA vaccines used in humans. The vaccines are well tolerated, safe, and highly efficacious; however, post-marketing surveillance is revealing potential rare adverse effects. We report a case of symptomatic acute myocarditis following administration of the second dose of mRNA-1273 SARS-CoV-2 vaccine. CASE SUMMARY: A 44-year-old man presented with chest pain and ST-segment elevation 4 days after receiving the second dose of mRNA-1273 SARS-CoV-2 Vaccine. Emergent coronary angiogram showed minimal coronary artery disease. Cardiac magnetic resonance imaging confirmed acute myocarditis. Diagnosis of vaccine-associated myocarditis was made given the temporal relationship and supportive treatment initiated. Follow-up at 1 month confirmed complete symptomatic recovery and echocardiogram demonstrated normalization of cardiac function. DISCUSSION: Acute myocarditis should be considered in patients who present with chest pain or dyspnoea within days of receiving mRNA-1273 SARS-CoV-2 vaccination, especially after the second dose. This may be managed successfully with supportive therapies with complete recovery of cardiac function and symptoms. Further research is warranted to determine the mechanisms by which mRNA vaccines may cause myocarditis and for potential long-term cardiovascular injury.

Ocular Toxicity of Targeted Anticancer Agents.

The proliferation of targeted anticancer agents over the last two decades has revolutionized cancer treatment and improved survival in many previously refractory malignancies. However, many agents are associated with characteristic ophthalmic adverse effects. It is important that ophthalmologists recognize and maintain a high index of suspicion for these side effects in patients on targeted therapy. Most ophthalmic adverse effects can be treated with specific ocular therapy without discontinuation of cancer treatment, although it is important to be aware of the life-threatening and vision-threatening circumstances that would require therapy cessation in conjunction with the patient's oncologist. This review aims to summarize the ophthalmic adverse effects of targeted and hormonal anticancer agents and briefly describe their management.

Diagnostic and prognostic biomarker potential of kallikrein family genes in different cancer types.

PURPOSE: The aim of this study was to compare and contrast the expression of all members of the Kallikrein (KLK) family of genes across 15 cancer types and to evaluate their utility as diagnostic and prognostic biomarkers. RESULTS: Severe alterations were found in the expression of different Kallikrein genes across various cancers. Interestingly, renal clear cell and papillary carcinomas have similar kallikrein expression profiles, whereas, chromophobe renal cell carcinoma has a unique expression profile. Several KLK genes have excellent biomarker potential (AUC > 0.90) for chromophobe renal cell carcinoma (KLK2, KLK3, KLK4, KLK7, KLK15), renal papillary carcinoma (KLK1, KLK6, KLK7), clear cell renal cell carcinoma (KLK1, KLK6), thyroid carcinoma (KLK2, KLK4, KLK13, KLK15) and colon adenocarcinoma (KLK6, KLK7, KLK8, KLK10). Several KLK genes were significantly associated with mortality in clear cell renal cell carcinoma (KLK2: HR = 1.69; KLK4: HR = 1.63; KLK8: HR = 1.71; KLK10: HR = 2.12; KLK11: HR = 1.76; KLK14: HR = 1.86), papillary renal cell carcinoma (KLK6: HR = 3.38, KLK7: HR = 2.50), urothelial bladder carcinoma (KLK5: HR = 1.89, KLK6: HR = 1.71, KLK8: HR = 1.60), and hepatocellular carcinoma (KLK13: HR = 1.75). METHODS: The RNA-seq gene expression data were downloaded from The Cancer Genome Atlas (TCGA). Statistical analyses, including differential expression analysis, receiver operating characteristic curves and survival analysis (Cox proportional-hazards regression models) were performed. CONCLUSIONS: A comprehensive analysis revealed the changes in the expression of different KLK genes associated with specific cancers and highlighted their potential as a diagnostic and prognostic tool.