Development and international validation of custom-engineered and code-free deep-learning models for detection of plus disease in retinopathy of prematurity: a retrospective study.

BACKGROUND: Retinopathy of prematurity (ROP), a leading cause of childhood blindness, is diagnosed through interval screening by paediatric ophthalmologists. However, improved survival of premature neonates coupled with a scarcity of available experts has raised concerns about the sustainability of this approach. We aimed to develop bespoke and code-free deep learning-based classifiers for plus disease, a hallmark of ROP, in an ethnically diverse population in London, UK, and externally validate them in ethnically, geographically, and socioeconomically diverse populations in four countries and three continents. Code-free deep learning is not reliant on the availability of expertly trained data scientists, thus being of particular potential benefit for low resource health-care settings. METHODS: This retrospective cohort study used retinal images from 1370 neonates admitted to a neonatal unit at Homerton University Hospital NHS Foundation Trust, London, UK, between 2008 and 2018. Images were acquired using a Retcam Version 2 device (Natus Medical, Pleasanton, CA, USA) on all babies who were either born at less than 32 weeks gestational age or had a birthweight of less than 1501 g. Each images was graded by two junior ophthalmologists with disagreements adjudicated by a senior paediatric ophthalmologist. Bespoke and code-free deep learning models (CFDL) were developed for the discrimination of healthy, pre-plus disease, and plus disease. Performance was assessed internally on 200 images with the majority vote of three senior paediatric ophthalmologists as the reference standard. External validation was on 338 retinal images from four separate datasets from the USA, Brazil, and Egypt with images derived from Retcam and the 3nethra neo device (Forus Health, Bengaluru, India). FINDINGS: Of the 7414 retinal images in the original dataset, 6141 images were used in the final development dataset. For the discrimination of healthy versus pre-plus or plus disease, the bespoke model had an area under the curve (AUC) of 0·986 (95% CI 0·973-0·996) and the CFDL model had an AUC of 0·989 (0·979-0·997) on the internal test set. Both models generalised well to external validation test sets acquired using the Retcam for discriminating healthy from pre-plus or plus disease (bespoke range was 0·975-1·000 and CFDL range was 0·969-0·995). The CFDL model was inferior to the bespoke model on discriminating pre-plus disease from healthy or plus disease in the USA dataset (CFDL 0·808 [95% CI 0·671-0·909, bespoke 0·942 [0·892-0·982]], p=0·0070). Performance also reduced when tested on the 3nethra neo imaging device (CFDL 0·865 [0·742-0·965] and bespoke 0·891 [0·783-0·977]). INTERPRETATION: Both bespoke and CFDL models conferred similar performance to senior paediatric ophthalmologists for discriminating healthy retinal images from ones with features of pre-plus or plus disease; however, CFDL models might generalise less well when considering minority classes. Care should be taken when testing on data acquired using alternative imaging devices from that used for the development dataset. Our study justifies further validation of plus disease classifiers in ROP screening and supports a potential role for code-free approaches to help prevent blindness in vulnerable neonates. FUNDING: National Institute for Health Research Biomedical Research Centre based at Moorfields Eye Hospital NHS Foundation Trust and the University College London Institute of Ophthalmology. TRANSLATIONS: For the Portuguese and Arabic translations of the abstract see Supplementary Materials section.

International collaboration for the development of clinical guidelines in low and middle-income countries: case study on the development of a national framework and clinical guidelines for diabetic retinopathy in Ghana.

BACKGROUND: Diabetic retinopathy is a leading cause of blindness in many countries across the world. Ghana has seen a rise in diabetic retinopathy and is working on various strategies to prevent blindness. Clinical guidelines are seen as a promising strategy for improving quality and reducing cost of care. Little is known about the processes of collaborative guideline development in the African context. METHODS: This case study discusses the process of developing clinical guidelines for diabetic retinopathy in Ghana via a collaboration with the Kenya team that had previously developed guidelines for Kenya. RESULTS: The main lesson learnt was the ability to overcome challenges. The main output achieved was the draft national framework, guidelines and training slides on the guidelines. CONCLUSION: Horizontal international collaboration can aid development of clinical guidelines.

Use of ziv-aflibercept in diabetic macular edema in a Ghanaian population.

AIM: To investigate the use of intravitreal ziv-aflibercept (IVZ) in Ghanaian patients with diabetic macular edema (DME). METHODS: A retrospective study of patients with DME, who had been treated with IVZ (1.25 mg/0.05 ml), as part of routine clinical practice, on pro re nata basis between 2016 and 2018 who had a minimum follow-up of 6 months was retrieved and analyzed. The primary outcome measure was change in best-corrected visual acuity (BCVA) at 6 months. Secondary outcome measures are change in BCVA at 12 months and at the last follow-up visit, adverse events and change in central macular thickness (CMT). RESULTS: Twenty-five eyes of 17 patients (11 males) were included in this study. Their mean age was 60.82 ± 7.70 years and the mean duration of follow-up was 9.52 ± 3.31 months. The mean baseline BCVA (logMAR) of 0.65 ± 0.3 improved to 0.34 ± 0.16 (p < 0.0001) and 0.22 ± 0.15 (p = 0.0004) at 6 and 12 months, respectively. Twelve (48%) eyes had a visual gain of at least three lines at 6 months and 4 of 12 eyes (33.3%) at 1 year. There was a significant reduction in the mean CMT at 6 and 12 months and at the last follow-up visit compared to baseline (p < 0.0001). The adverse events recorded were raised intraocular pressure (four eyes) at 3, 6, and 12 months post injection, increased blood pressure in a patient with known systemic hypertension and transient memory loss in one patient. CONCLUSION: IVZ (1.25 mg) was associated with significant improvement in BCVA and reduction in CMT at 6 and 12 months in eyes with DME. A randomized clinical trial is warranted to assess this potentially cost-effective intervention for DME in low-resource settings.

Efficacy of intravitreal ziv-aflibercept in patients with macular edema following retinal vein occlusion in Korle-Bu Teaching Hospital, Ghana: a retrospective case series.

AIM: To evaluate the efficacy of ziv-aflibercept in Ghanaian patients with macular edema (ME) secondary to retinal vein occlusion (RVO). METHODOLOGY: In this retrospective study, the medical records of patients with ME secondary to RVO who had been treated with intravitreal ziv-aflibercept (IVZ) (1.25 mg/0.05 ml), as part of routine clinical practice, on pro re nata basis with a minimum follow-up of 6 months were retrieved and analyzed. The main outcome measures are mean change in best-corrected visual acuity (BCVA) and central subfield foveal thickness (CSFT) measured on optical coherence tomography from baseline to 12 months post-IVZ, and ocular and systemic safety. RESULTS: Forty-three eyes were included in this study. Their mean age was 62.8 ± 11.9 years, 67.4% had at least 12-month duration of follow-up, 50% had primary open-angle glaucoma and 38 (88.4%) eyes were treatment naive. There was significant improvement in mean BCVA in LogMAR at 1 month post-initiation of IVZ (0.8 ± 0.5 vs. 1.1 ± 0.6), and visual improvement was maintained up to 12 months (p < 0.001). Eyes with ME following BRVO had better mean BCVA at baseline and on subsequent visits compared to eyes with CRVO/HRVO (p = 0.01). There was significant reduction in mean CSFT up to 12 months post-IVZ injection compared to baseline (p < 0.001). Ocular complications observed were consistent with complications associated with RVO. CONCLUSION: We have observed significant improvement in functional and anatomic outcomes 12 months post-initiation of IVZ. There is the need to confirm long-term efficacy and safety of IVZ in a large prospective study.

Incidence and risk factors of retinopathy of prematurity in Korle-Bu Teaching Hospital: a baseline prospective study.

OBJECTIVE: To determine the incidence of retinopathy of prematurity (ROP) and any associated risk factors among preterm infants at the Neonatal Intensive Care Unit (NICU) of Korle-Bu Teaching Hospital (KBTH). DESIGN: Prospective study. SETTING: Level 3 NICU of KBTH from June 2018 to February 2019. PARTICIPANTS: Eligible infants with birth weight (BW) less than 2 kg or gestational age (GA) less than 37 weeks were examined at scheduled intervals until full maturity of their retina. OUTCOME MEASURES: The primary outcome measure was cumulative incidence of ROP and secondary outcome measure was risk factors associated with ROP. RESULTS: Of the 401 infants, 222 were females (55.4%), mean±SD GA was 32.3±2.4 weeks (median 32, IQR 31 to 34) and mean BW 1.6±0.4 kg (median 1.5, IQR 1.3 to 1.9). The cumulative incidence of ROP was 13.7% (95% CI: 10.5 to 17.5%), with 1.8% (seven infants) having type 1 ROP. Increased risk of ROP was observed in babies with supplemental oxygen exposure (p<0.001), BW less than 1.5 kg (p=0.019), confirmed neonatal sepsis (p=0.001), nasogastric tube feeding (p=0.03) and poor pupillary dilation (0.032). A reduced risk of ROP was observed in boys (p=0.004) and after delivery by caesarean section (p=0.019). CONCLUSION: The rates of ROP at KBTH are comparable to other NICUs in sub-Saharan Africa. Birth weight less than 1.5 kg, confirmed neonatal sepsis, nasogastric tube feeding and poor pupil dilation were independently associated with increased incidence of ROP. ROP screening should be a part of the routine service for premature infants in Ghana.

ENTEROCOCCUS ENDOPHTHALMITIS: Clinical Settings, Antimicrobial Susceptibility, and Management Outcomes.

PURPOSE: To report the clinical presentation and management outcome of patients with endophthalmitis caused by Enterococcus species and to report the susceptibility profile of the isolates. METHODS: Twenty-nine cases with culture-proven Enterococcus endophthalmitis from January 2005 to May 2018 underwent vitrectomy/vitreous biopsy, intravitreal antibiotic with or without additional procedures. The undiluted vitreous was subjected to microbiologic evaluation. A favorable anatomical outcome was defined as preservation of the globe, absence of hypotony, attached retina, and absence of active inflammation at the last visit. Favorable visual outcome was final visual acuity ≥20/400. RESULTS: There were 24 men (82.8%). Mean age at presentation was 32.89 ± 25.25 years (median 24 years). Inciting event was open globe injury in 18 (62%), endogenous in 5 (17.24%), postcataract surgery in 3 (10.34%), postscleral buckling in 2 (6.89%), and postkeratoplasty in 1 (3.44%). Enterococcus casseliflavus was the commonest species isolated (14/29, 48.27%) followed by E. faecalis (9/29, 31%). Susceptibility to vancomycin was seen in 27/29 isolates (93%). Visual acuity was ≤20/400 in all eyes at presentation and ≥20/400 in 10/29 cases (34.48%) at final visit. Anatomical success was seen in 18/29 eyes (62%). Corneal involvement was high at 24/29 eyes (82.75%). CONCLUSION: Enterococcus is not an uncommon organism in the setting of endophthalmitis after open globe injury. Resistance to vancomycin is rare. Multidrug resistance pattern is restricted to E. faecalis. Visual outcome is poor despite early and appropriate therapy due to inherent organism virulence.

Intravitreal Ziv-Aflibercept : Safety Analysis in Eyes Receiving More Than Ten Intravitreal Injections.

Purpose: To report the ocular and systemic adverse events in eyes receiving ≥10 intravitreal ziv-aflibercept (IVZ) injections.Methods: Medical records of patients who received ≥10 IVZ for various chorioretinal conditions with minimum follow up period of 12 months were retrospectively analysed. These eyes received standard dose of IVZ (1.25 mg/0.05 ml) on pro-re-nata (PRN) or treat and extend (T&E) protocol. The primary study outcome was ocular and systemic adverse events related to IVZ injections whereas secondary outcomes were change in best corrected visual acuity (BCVA) and central macular thickness (CMT) on optical coherence tomography (OCT) at last visit compared to baseline. Comparison of BCVA and CMT at baseline and final visit was done using paired t-test.Results: A total of 94 eyes which received a mean ± standard deviation (mean±SD = 14.4 ± 4.6) IVZ injections were studied. A total of 41 eyes were treatment naïve whereas 53 eyes received intravitreal injections in the past with last injection at least 3 months prior. Mean (±SD) follow up period was 26.7 ± 8.7 months. Ocular adverse events were limited with a case each of acute iridocyclitis, endophthalmitis, cataract progression and early epiretinal membrane formation. No systemic events were recorded within a month of IVZ injection. There was a significant improvement in BCVA (p = 0.001) and change in CMT (p = 0.001) at last visit.Conclusion: Ocular use of ziv-aflibercept is safe with limited ocular and systemic side effects. Multiple injections of IVZ can be used in various chorioretinal diseases over the long term.

Safety of intravitreal ziv-aflibercept in choroido-retinal vascular diseases: A randomised double-blind intervention study.

AIM: To evaluate the safety of 1.25mg and 2mg intravitreal ziv-aflibercept (IVZ) in Ghanaian eyes with choroido-retinal vascular diseases. DESIGN: Prospective, randomised, double blind, interventional study. METHODS: Twenty patients with centre involving macular oedema in diabetic retinopathy, retinal vein occlusion, and neovascular age-related macular degeneration were assigned to 2 groups receiving 3 doses of 1.25mg/0.05ml (group 1) and 2mg/0.08ml IVZ (Group 2) at 4 weekly intervals. Safety data was collected after 30 minutes, 1 and 7 days, and 4, 8 and 12 weeks after injection. Changes in continuous variables were compared using paired t-test and categorical variables were compared using chi-square test of proportions. Repeated-Measures ANOVA with nesting test was used to compare variations in continuous variables by IVZ dose over time. Primary outcome measures were ocular and systemic adverse events at 4 weeks. RESULTS: Eleven females and nine males, with mean age of 63.2± 7.3 years were included. Ocular adverse events included subconjunctival haemorrhage in 1 eye, intraocular pressure (IOP) >21mmHg at 30 minutes in 6 eyes and mild pain in 3 eyes at 1-day. There was no significant difference in IOP rise between the 2 groups at 30 minutes (p = 0.21). No other ocular or systemic adverse events were observed. There was significant improvement in the best corrected visual acuity (LogMAR) from 0.95±0.6 to 0.6±0.4 (p<0.01) and 0.47±0.3 (p<0.01), reduction in central subfield foveal thickness from 405.9±140 um at baseline to 255.6±75 um (p<0.01) and 238±88 um (p<0.01) at 4 and 12 weeks respectively, although no difference was observed between the 2 groups (p = 0.34). CONCLUSION: IVZ at 1.25mg and 2mg had similar safety profiles, and did not have any major unexpected adverse events. Further studies with larger cohorts are required to confirm efficacy.

Safety of 5914 intravitreal ziv-aflibercept injections.

PURPOSE: To analyse the pooled safety data of intravitreal ziv-aflibercept (IVZ) therapy for various retinal conditions. METHODS: This was a retrospective, observational study which included patients from 14 participating centres who received IVZ. The medical records of patients who received IVZ from March 2015 through October 2017 were evaluated. Patient demographics and ocular details were compiled. Ocular and systemic adverse events that occurred within 1 month of IVZ injections were recorded and defined as either procedure-related or drug-related. RESULTS: A total of 1704 eyes of 1562 patients received 5914 IVZ injections (mean±SD: 3.73±3.94) during a period of 2.5 years. The age of patients was 60.6±12.8 years (mean±SD) and included diverse chorioretinal pathologies. Both ocular (one case of endophthalmitis, three cases of intraocular inflammation, and one case each of conjunctival thinning/necrosis and scleral nodule) and systemic adverse events (two cases of myocardial infarction, one case of stroke and two deaths) were infrequent. CONCLUSION: This constitutes the largest pooled safety report on IVZ use and includes patients from 14 centres distributed across the globe. It shows that IVZ has an acceptable ocular and systemic safety profile with incidences of adverse events similar to those of other vascular endothelial growth factor inhibitory drugs. The analysis supports the continued use of IVZ in various retinal disorders.

Focal Choroidal Excavation in Retinal Dystrophies.

AIM: To investigate the presence of focal choroidal excavation (FCE) in patients with retinitis pigmentosa (RP), Stargardt's disease (STGD), and Best disease in the Indian population. METHODS: This retrospective consecutive case series included 309 eyes of 157 patients with RP (183 eyes), STGD (93 eyes), and Best disease (33 eyes) with good-quality, enhanced-depth spectral domain optical coherence tomography scans. Comprehensive ophthalmic examination data were collected. Characteristics of FCE, including location of FCE, type (conforming and non-conforming), maximal width, and depth, were noted. RESULTS: FCE was found in 2 out of 33 (6%) eyes with Best disease and no FCE was found in eyes with RP or STGD. The location of the FCE was extrafoveal in both cases. The first case had non-conforming FCE while the second case had the conforming type and the FCE occurred in association with choroidal neovascularization in the second case. The first case maintained good visual acuity of 20/20 over the entire period of follow-up (14 months), while the second case had a visual acuity of 20/200 at the last follow-up (three years) due to scarred choroidal neovascular membranes. The FCE showed no change in both eyes over the entire duration of follow-up. CONCLUSION: Focal choroidal excavation was found in 6% of eyes with Best disease, which remained stable throughout follow up. Eyes with RP and STGD did not have any FCE. Further studies are required to determine the role of vitelliform material in FCE development in Best disease.

One-year outcome of intravitreal ziv-aflibercept therapy for non-responsive neovascular age-related macular degeneration.

AIM: To evaluate 12-month outcome of intravitreal ziv-aflibercept (IVZ) therapy in eyes with neovascular age-related macular degeneration (nAMD) that are non-responsive to bevacizumab and ranibizumab. METHODS: This retrospective study included 16 eyes (14 patients) with nAMD who were on prior treatment with bevacizumab and ranibizumab and were treated with as-needed IVZ (1.25 mg/0.05 mL) for 12 months. The primary outcome measure was the mean change in best corrected visual acuity (BCVA) and secondary outcome measures included mean change in central macular thickness (CMT), retinal pigment epithelial detachment (RPED) heights, longest treatment free interval, presence of subretinal fluid (SRF) and intraretinal fluid (IRF) and adverse events. RESULTS: There was no change in the mean logarithm of minimum angle of resolution (logMAR) BCVA at baseline and following treatment with IVZ therapy (p=0.978). The mean number of IVZ injections during 12 months was 5.9±3.3, and the mean number of antivascular endothelial growth factors (VEGFs) injections prior to switching to IVZ was 8.4±4.7. The mean treatment free interval was longer during IVZ therapy (114.4±67.1 days) compared with 76.3±54.6 days before IVZ therapy (p=0.03). Five (31.25%) eyes had visual gains of at least 0.1 logMAR, 3 (18.75%) eyes had stable BCVA (within 0.1 logMAR) and 8 (50%) eyes had BCVA decline of at least 0.1 logMAR. There was no significant difference in the mean CMT, RPED heights and presence of IRF and SRF at 12 months compared with baseline. No adverse events were noted. CONCLUSION: IVZ increased the treatment free interval in non-responders but no significant change in visual and anatomic outcomes.

Intravitreal ziv-aflibercept for the treatment of choroidal neovascularisation associated with conditions other than age-related macular degeneration.

AIM: To report the short-term outcomes of eyes with choroidal neovascularisation (CNV) associated with causes other than age-related macular degeneration (AMD) after treatment with intravitreal ziv-aflibercept (IVZ) injections. METHODS: This retrospective study included eyes with non-AMD-related CNV that were treated with IVZ (1.25 mg/0.05 mL) on a pro re nata basis. The primary outcome measure is the mean change in best-corrected visual acuity (BCVA) and secondary outcome measures include the mean change in central macular thickness (CMT) and adverse events. RESULTS: 23 eyes of 19 patients with CNV due to high myopia (9), macular telangiectasia (4), central serous chorioretinopathy (3), choroidal osteoma (2), choroiditis (2), Best's disease (2) and idiopathic (1) were treated. The mean follow-up period was 4±1.9 months. The median number of IVZ injections was 1 (range, 1-3) and the median treatment-free interval at the time of the final visit was 3 months (range, 1-8). The mean BCVA improved from 0.67 LogMAR to 0.58 LogMAR (p=0.0507). Nine of 23 (39%) eyes had BCVA gains of at least 0.1 LogMAR, 11 (48%) eyes had stable BCVA (within 0.1 LogMAR of baseline) and 3 (13%) eyes had a BCVA decline of at least 0.1 LogMAR at the final visit. The mean CMT improved significantly from baseline until the final visit (22 vs 174.5 µm; p=0.037). No ocular or systemic adverse events were noted. CONCLUSIONS: IVZ improves CMT in patients with CNV associated with causes other than AMD, but improvements in BCVA are modest.