Expert-Level Detection of Referable Glaucoma from Fundus Photographs in a Safety Net Population: The AI and Teleophthalmology in Los Angeles Initiative.

Purpose: To develop and test a deep learning (DL) algorithm for detecting referable glaucoma in the Los Angeles County (LAC) Department of Health Services (DHS) teleretinal screening program. Methods: Fundus photographs and patient-level labels of referable glaucoma (defined as cup-to-disc ratio [CDR] ≥ 0.6) provided by 21 trained optometrist graders were obtained from the LAC DHS teleretinal screening program. A DL algorithm based on the VGG-19 architecture was trained using patient-level labels generalized to images from both eyes. Area under the receiver operating curve (AUC), sensitivity, and specificity were calculated to assess algorithm performance using an independent test set that was also graded by 13 clinicians with one to 15 years of experience. Algorithm performance was tested using reference labels provided by either LAC DHS optometrists or an expert panel of 3 glaucoma specialists. Results: 12,098 images from 5,616 patients (2,086 referable glaucoma, 3,530 non-glaucoma) were used to train the DL algorithm. In this dataset, mean age was 56.8 ± 10.5 years with 54.8% females and 68.2% Latinos, 8.9% Blacks, 2.7% Caucasians, and 6.0% Asians. 1,000 images from 500 patients (250 referable glaucoma, 250 non-glaucoma) with similar demographics (p ≥ 0.57) were used to test the DL algorithm. Algorithm performance matched or exceeded that of all independent clinician graders in detecting patient-level referable glaucoma based on LAC DHS optometrist (AUC = 0.92) or expert panel (AUC = 0.93) reference labels. Clinician grader sensitivity (range: 0.33-0.99) and specificity (range: 0.68-0.98) ranged widely and did not correlate with years of experience (p ≥ 0.49). Algorithm performance (AUC = 0.93) also matched or exceeded the sensitivity (range: 0.78-1.00) and specificity (range: 0.32-0.87) of 6 LAC DHS optometrists in the subsets of the test dataset they graded based on expert panel reference labels. Conclusions: A DL algorithm for detecting referable glaucoma developed using patient-level data provided by trained LAC DHS optometrists approximates or exceeds performance by ophthalmologists and optometrists, who exhibit variable sensitivity and specificity unrelated to experience level. Implementation of this algorithm in screening workflows could help reallocate eye care resources and provide more reproducible and timely glaucoma care.

Racial and Sociodemographic Disparities in the Detection of Narrow Angles before Detection of Primary Angle-Closure Glaucoma in the United States.

PURPOSE: To assess the proportion of newly diagnosed cases of primary angle-closure glaucoma (PACG) with and without prior diagnosis of anatomical narrow angle (ANA) and to identify sociodemographic risk factors for late detection (PACG without prior ANA diagnosis). DESIGN: Retrospective cohort study. METHODS: One hundred two thousand six hundred seventeen patients with PACG were identified from the Optum Clinformatics Data Mart Database (2007-2019). Patients with newly diagnosed PACG met the following criteria: (1) diagnosis made by an ophthalmologist, (2) disease observable for at least 12 months before diagnosis, and (3) no history of treatment before diagnosis unless preceded by a diagnosis of ANA. Multivariate logistic regression modeling was performed to identify sociodemographic risk factors for late detection. MAIN OUTCOME MEASURES: Proportion of patients with newly diagnosed PACG without prior ANA diagnosis and sociodemographic factors associated with late detection. RESULTS: Thirty-one thousand forty-four patients were eligible. More than 70% of PACG cases were detected without prior ANA diagnosis, regardless of patient age, sex, or race. The odds of late detection were significantly higher (P < 0.001) among men (odds ratio [OR], 1.32; 95% confidence interval [CI], 1.25-1.40), Black patients (OR, 1.25; 95% CI, 1.15-1.37), and patients 80 years of age or older (OR, 1.28; 95% CI, 1.11-1.47) or living in Southern (OR, 1.30; 95% CI, 1.22-1.40) or Pacific (OR, 1.27; 95% CI, 1.16-1.36) regions. Findings were similar for patients with PACG with a record of gonioscopy and treatment or with a 24-month lookback period. CONCLUSIONS: Most patients who receive a new diagnosis of PACG in the United States do not have a prior diagnosis of ANA. The elderly, men, and Black patients are at higher risk of late detection. A need exists for increased disease awareness among providers and more accessible tools to detect patients at risk of developing PACG.

Start codon disruption with CRISPR/Cas9 prevents murine Fuchs' endothelial corneal dystrophy.

A missense mutation of collagen type VIII alpha 2 chain (COL8A2) gene leads to early-onset Fuchs' endothelial corneal dystrophy (FECD), which progressively impairs vision through the loss of corneal endothelial cells. We demonstrate that CRISPR/Cas9-based postnatal gene editing achieves structural and functional rescue in a mouse model of FECD. A single intraocular injection of an adenovirus encoding both the Cas9 gene and guide RNA (Ad-Cas9-Col8a2gRNA) efficiently knocked down mutant COL8A2 expression in corneal endothelial cells, prevented endothelial cell loss, and rescued corneal endothelium pumping function in adult Col8a2 mutant mice. There were no adverse sequelae on histology or electroretinography. Col8a2 start codon disruption represents a non-surgical strategy to prevent vision loss in early-onset FECD. As this demonstrates the ability of Ad-Cas9-gRNA to restore the phenotype in adult post-mitotic cells, this method may be widely applicable to adult-onset diseases, even in tissues affected with disorders of non-reproducing cells.

Dupilumab-Associated Ocular Surface Disease: Clinical Characteristics, Treatment, and Follow-Up.

PURPOSE: A consecutive case series of patients with dupilumab-associated ocular surface disease (DAOSD) that describes common ocular symptoms and signs, proposes a symptom disease severity grading system, and describes treatment strategies of DAOSD patients was evaluated. METHODS: A retrospective chart review of patients with concomitant dupilumab-treated atopic dermatitis and DAOSD with ophthalmic evaluation between January 2014 and May 2019 was conducted. RESULTS: Twenty-nine patients (mean age 46 years, M/F: 12/17) with 57 ophthalmic exams were identified. The most common ocular symptoms included irritation/pain (n = 28, 97%), redness (n = 24, 83%), pruritus (n = 18, 62%), discharge (n = 18, 62%), and light sensitivity (n = 6, 21%). The most frequent signs included conjunctival injection (n = 18, 62%), superficial punctate keratitis (n = 16, 55%), and papillary reaction (n = 8, 28%). Topical corticosteroids (TCS) (n = 23, 79%), tacrolimus (n = 6, 21%), and artificial tears (n = 7, 24%) were the most commonly used therapies. Of those with follow-up documentation (n = 21), 20 were noted to have partial or complete response with TCS based on symptoms and reduction of signs. Using our proposed symptom-based grading scale, scaled 1 to 5 based on the presence of common symptoms listed above, 66% (n = 19) requiring topical immunomodulating therapy were found in the 'severe' group (≥3 symptoms) and 17% (n = 5) were found in the 'mild' group (≤2 symptoms). CONCLUSIONS: This study provides insight into the commonly presenting ocular signs and symptoms associated with DAOSD and highlights the efficacy of TCS and other immunomodulators in improving symptoms associated with DAOSD. Based on our findings, we propose a symptom-based grading system that can guide nonophthalmic physicians regarding ophthalmology consult.

Effects on phacoemulsification efficiency and chatter at variable longitudinal ultrasound settings when combined with constant torsional energy.

PURPOSE: To evaluate longitudinal power settings for optimally efficient lens fragment removal, using the Centurion machine. SETTING: John A. Moran Eye Center, University of Utah, Salt Lake City, USA. DESIGN: Experimental study. METHODS: Porcine lens nuclei were cut into 2.0 mm cubes. Experiments were conducted at 100% torsional power; vacuum set at 500 mm Hg, aspiration 50 mL/min, and intraocular pressure 110 mm Hg. A 20-degree tip with a 30-degree bevel was used. Longitudinal power was tested between 20% and 100%. Efficiency (time for fragment removal) and chatter (the number of times the fragment bounced from the tip) were measured. RESULTS: A linear increase in efficiency was observed from 20% to 100% longitudinal power (R = 0.9281, slope = -0.0271). An efficiency slope change occurred at 60% power, with the largest incremental change in efficiency between 20% and 60% (R = 0.9756, slope = -0.0394) and a lesser change between 60% and 100% (R = 0.9827, slope = -0.0121). Chatter analysis showed minimal events at 20% to 60%, but a significant increase at >80% (P = .005). This increase appeared to be incremental (R = 0.8929). CONCLUSIONS: Increasing longitudinal power, with all other settings constant, increased efficiency. Greatest efficiency gains were observed between 20% and 60%. At 80% and 100%, chatter events increased significantly. With a goal of recommending optimally efficient settings while minimizing excess energy and chatter, adding 60% of longitudinal power to 100% torsional power was shown to be the best setting to increase efficiency and avoid repulsion in these vacuum and aspiration settings.

Optimum on-time duty cycle for a transversal ultrasound machine.

PURPOSE: To evaluate the optimum on-time setting for the most efficient removal of lens fragments using micropulse ultrasound (US) and Ellips FX transversal US in the Whitestar Signature Pro phacoemulsification machine. SETTING: John A. Moran Eye Center Laboratories, University of Utah, Salt Lake City, Utah, USA. DESIGN: Experimental study. METHODS: Porcine lens nuclei were soaked in formalin for 2 hours and cut into 2.0 mm cubes. The US machine was used with a bent 0.9 mm phaco tip and a 30-degree bevel. The off time was set to 6 milliseconds (ms) and the on time varied from 4 to 10 ms in 1 ms increments. Efficiency (time for fragment removal) and chatter (number of times the fragment bounced from the tip) were measured. RESULTS: A linear incremental increase in efficiency was observed between 4 ms and 6 ms. The most statistically significant efficiency was achieved with an on time of 6 ms. On times shorter than 6 ms were significantly less efficient (P = .05). Greater on times (7 to 10 ms) did not result in a significant difference in efficiency (P = .72), but did appear to have more chatter events when comparing on-time settings of 7 to 10 ms with 4 to 6 ms (P = .02). CONCLUSIONS: With micropulse transversal US, 6 ms of on time was as efficient as longer on times. To maximize phacoemulsification efficiency and minimize chatter events, an on time of 6 ms is recommended.

The Effect of Pulsing on Transverse Ultrasound Efficiency and Chatter.

PURPOSE: To evaluate the effects of micropulse, long pulse, and continuous ultrasound on transverse ultrasound using Abbott Medical Optics' (AMO) WhiteStar Signature Pro with the Ellips FX handpiece. DESIGN: In vitro laboratory study. METHODS: This study was conducted at the John A. Moran Eye Center Laboratory, University of Utah, Salt Lake City, Utah, USA. Porcine lenses were hardened in formalin for 2 hours and equilibrated in basic salt solution (BSS) over a 24-hour period. The lenses were then cubed in 2.0 × 2.0-mm pieces. These pieces were stored in BSS until the time of experimentation. The AMO WhiteStar Signature Pro machine (Abbott Medical Optics) with the Ellips FX handpiece and a 0.9-mm bent Dewey tip with a 30-degree bevel (Microsurgical Technology Inc) were used for phacoemulsification. Three runs of 20 lenses each were performed, measuring efficiency and chatter. Transverse ultrasound varied in the 3 runs and included continuous, 6 ms on/off micropulse, and 50 ms on/off long pulse. RESULTS: Micropulse was more efficient than long pulse by 43% (P = .00003) and continuous by 42% (P = .000387). There were also less chatter events with micropulse than with long-pulse and continuous ultrasound. However, this difference did not reach significance. CONCLUSION: The 6 ms on and 6 ms off micropulse transverse 3-dimensional ultrasound is more efficient and produces fewer chatter events than both long-pulse and continuous ultrasound.

Targeted Intraceptor Nanoparticle for Neovascular Macular Degeneration: Preclinical Dose Optimization and Toxicology Assessment.

Neovascular age-related macular degeneration (AMD) is treated with anti-VEGF intravitreal injections, which can cause geographic atrophy, infection, and retinal fibrosis. To minimize these toxicities, we developed a nanoparticle delivery system for recombinant Flt23k intraceptor plasmid (RGD.Flt23k.NP) to suppress VEGF intracellularly within choroidal neovascular (CNV) lesions in a laser-induced CNV mouse model through intravenous administration. In the current study, we examined the efficacy and safety of RGD.Flt23k.NP in mice. The effect of various doses was determined using fluorescein angiography and optical coherence tomography to evaluate CNV leakage and volume. Efficacy was determined by the rate of inhibition of CNV volume at 2 weeks post-treatment. RGD.Flt23k.NP had peak efficacy at a dose range of 30-60 µg pFlt23k/mouse. Using the lower dose (30 µg pFlt23k/mouse), RGD.Flt23k.NP safety was determined both in single-dose groups and in repeat-dose (three times) groups by measuring body weight, organ weight, hemoglobin levels, complement C3 levels, and histological changes in vital organs. Neither toxicity nor inflammation from RGD.Flt23k.NP was detected. No side effect was detected on visual function. Thus, systemic RGD.Flt23k.NP may be an alternative to standard intravitreal anti-VEGF therapy for the treatment of neovascular AMD.

AAV2 delivery of Flt23k intraceptors inhibits murine choroidal neovascularization.

Long-term inhibition of extracellular vascular endothelial growth factor (VEGF) in the treatment of age-related macular degeneration (AMD) may induce retinal neuronal toxicity and risk other side effects. We developed a novel strategy which inhibits retinal pigment epithelium (RPE)-derived VEGF, sparing other highly sensitive retinal tissues. Flt23k, an intraceptor inhibitor of VEGF, was able to inhibit VEGF in vitro. Adeno-associated virus type 2 (AAV2)-mediated expression of Flt23k was maintained for up to 6 months postsubretinal injection in mice. Flt23k was able to effectively inhibit laser-induced murine choroidal neovascularization (CNV). VEGF levels in the RPE/choroid complex decreased significantly in AAV2.Flt23k treated eyes. Neither retinal structure detected by Heidelberg Spectralis nor function measured by electroretinography (ERG) was adversely affected by treatment with AAV2.Flt23k. Hence AAV2.Flt23k can effectively maintain long-term expression and inhibit laser-induced CNV in mice through downregulation of VEGF while maintaining a sound retinal safety profile. These findings suggest a promising novel approach for the treatment of CNV.

Pharmacokinetics and efficacy of Bioerodible Dexamethasone implant in Concanavalin A-induced uveitic cataract rabbit model.

PURPOSE: To advance therapy for the treatment of concurrent uveitis and post-cataract surgical inflammation; we evaluated pharmacokinetics and pharmacodynamics of Bioerodible Dexamethasone Implant (BDI) containing 0.3 mg of dexamethasone (DXM) in Concanavalin A (Con A) induced uveitis followed by phacoemulsification in New Zealand White (NZW) rabbits. METHODS: The BDI was implanted in the inferior fornix of the capsular bag after intravitreal injection of Con A and ensuing phacoemulsification in NZW rabbits; standard-of-care topical 0.1% dexamethasone drops served as control. DXM was quantified by liquid chromatography-tandem mass spectrometry and pharmacokinetics of DXM in disease vs. healthy eyes was compared. All eyes were assessed clinically using slit lamp biomicroscopy and Draize scoring scale. Retinal thickness and histological analyses were performed to evaluate retinal edema, inflammation and implant biocompatibility respectively. RESULTS: In Con A-induced inflammatory uveitic cataract model the BDI controlled anterior and posterior segment inflammation as well as retinal thickening more effectively than topical drops. The exposure (AUC0-t) of DXM with BDI is superior in all ocular tissues, while topical drops did not achieve therapeutic posterior segment levels and did not control inflammation nor prevent retinal edema and architectural disruption. CONCLUSIONS: Our results demonstrate the superiority of the BDI in suppressing Con A-induced inflammation and retinal edema in NZW rabbits and highlight the need for sustained bidirectional delivery of potent anti-inflammatory agents for 5 to 6 weeks to optimize clinical outcomes.