Rebecca L. Thomas

Aim: This study aimed to determine the effectiveness of deep learning models for detecting and predicting diabetes related kidney disease (DRKD) using two‐dimensional retinal images. Methods: A systematic review was conducted across eight databases from inception to July 14, 2026. Eligible cross‐sectional and longitudinal studies assessed deep learning models for DRKD detection or prediction using retinal images alone, clinical data alone, or hybrid models. Data on study design, model development, validation, and performance metrics were extracted, with study quality appraised using QUADAS‐2 and post hoc, PROBAST+AI. Results were synthesized narratively, with random effects meta analysis for hybrid models. Results: Of 367 studies, six met inclusion criteria. Three cross‐sectional studies of hybrid models contributed to the meta‐analysis for DRKD detection, yielding pooled AUC 0.88, sensitivity 0.80, and specificity 0.81, though heterogeneity was high (I2 93.7%–99.8%). Three further studies were synthesized narratively rather than pooled: two cross‐sectional studies used more severe or narrower CKD case definitions (AUC 0.86 and 0.96 respectively), and only one longitudinal study predicted the 5‐year risk of DRKD (AUC 0.81, sensitivity 0.70, and specificity 0.77). External validation showed hybrid models consistently outperformed retinal‐only (ΔAUC 0.03–0.11), but their advantage over clinical‐only models was inconsistent. Conclusion: Deep learning models may be useful for detecting and predicting DRKD from retinal images, with or without clinical data. However, the small number of heterogenous studies restricts generalizability. Future work should prioritize robust model development, and validation using large, diverse populations.

Background: Coronary heart disease (CHD) is the leading cause of death worldwide, constituting a growing health and social burden. People with cardiometabolic disorders are more likely to develop CHD. Retinal image analysis is a novel and noninvasive method to assess microvascular function. We aim to investigate whether retinal images can be used for CHD risk estimation for people with cardiometabolic disorders. Methods: We have conducted a case–control study at Shenzhen Traditional Chinese Medicine Hospital, where 188 CHD patients and 128 controls with cardiometabolic disorders were recruited. Retinal images were captured within two weeks of admission. The retinal characteristics were estimated by the automatic retinal imaging analysis (ARIA) algorithm. Risk estimation models were established for CHD patients using machine learning approaches. We divided CHD patients into a diabetes group and a non-diabetes group for sensitivity analysis. A ten-fold cross-validation method was used to validate the results. Results: The sensitivity and specificity were 81.3% and 88.3%, respectively, with an accuracy of 85.4% for CHD risk estimation. The risk estimation model for CHD with diabetes performed better than the model for CHD without diabetes. Conclusions: The ARIA algorithm can be used as a risk assessment tool for CHD for people with cardiometabolic disorders.

Background: Patient education is a fundamental aspect of self-management of diabetes. The aim of this study was to understand whether a social media platform is a viable method to deliver education to people with diabetes and understand if people would engage and interact with it. Methods: Education sessions were provided via 3 platforms in a variety of formats. “Tweetorials” and quizzes were delivered on the diabetes101 Twitter account, a virtual conference via Zoom and video presentations uploaded to YouTube. Audience engagement during and after the sessions were analyzed using social media metrics including impressions and engagement rate using Twitter analytics, Tweepsmap, and YouTube Studio. Results: A total of 22 “tweetorial” sessions and 5 quizzes with a total of 151 polls (both in tweetorial and quiz sessions) receiving a total of 21,269 votes took place. Overall, the 1-h tweetorial sessions gained 1,821,088 impressions with an engagement rate of 6.3%. The sessions received a total of 2,341 retweets, 2,467 replies and 10,060 likes. The quiz days included 113 polls receiving 16,069 votes. The conference covered 8 topics and was attended live by over 100 people on the day. The video presentations on YouTube have received a total of 2,916 views with a watch time of 281 h and 8,847 impressions. Conclusion: Despite the limitations of social media, it can be harnessed to provide relevant reliable information and education about diabetes allowing people the time and space to learn at their own pace.

Diabetic retinopathy is a neurovascular complication of diabetes which is an avoidable cause of sight loss. Since the introduction of screening for diabetic retinopathy in the UK along with improving treatments for both diabetes and diabetic retinopathy it is no longer the leading cause of blindness in the UK. Screening detects diabetic retinopathy at an early stage when optimisation of blood glucose, blood pressure and cholesterol are effective in reducing the progression of diabetic retinopathy. Once proliferative diabetic retinopathy or diabetic macular oedema develops treatment includes laser photocoagulation, intravitreal vascular endothelial growth factor inhibitor injections, steroid injection and vitrectomy. With screening to detect diabetic retinopathy, improved therapies and technologies to help people manage diabetes and treatments for diabetic retinopathy sight loss is preventable for the majority of people. Copyright © 2021 John Wiley & Sons.

Aims: To understand factors associated with repeat non‐attendance at screening for diabetes‐related retinopathy.Methods: Retrospective observational study using anonymised data from Diabetic Eye Screening Wales for people with a full history of screening invitations and attendances was linked with primary and secondary care records held in the Secure Anonymised Information Linkage Databank. Repeat non‐attendance was defined as no record of attendance during any 36‐month period despite three cycles of annual screening invitations. The associations between repeat non‐attendance and potential risk factors were examined using multivariable logistic regression analysis, stratified according to type 1 and type 2 diabetes.Results: A total of 18% with type 1 diabetes (1146/6513) and 8% with type 2 diabetes (12,475/156,525) were repeat non‐attenders. Participants attending their very first appointment were least likely to become repeat non‐attenders [odds ratio (95% confidence interval)]: type 1 diabetes: 0.12 (0.09, 0.17) and type 2 diabetes: 0.08 (0.07, 0.09). For both types of diabetes, those of a younger age, living in areas of higher deprivation and subject to multiple house moves were at greater risk of becoming repeat non‐attenders.Conclusion/interpretation: A more tailored approach is needed for the younger population, those living in areas of higher deprivation and/or undergoing multiple residential relocation and to ensure attendance at their initial appointment to minimise future repeat non‐attendance.

Aims/hypothesis: Few studies examine the association between age at diagnosis and subsequent complications from type 2 diabetes. This paper aims to summarise the risk of mortality, macrovascular complications and microvascular complications associated with age at diagnosis of type 2 diabetes. Methods: Data were sourced from MEDLINE and All EBM (Evidence Based Medicine) databases from inception to July 2018. Observational studies, investigating the effect of age at diabetes diagnosis on macrovascular and microvascular diabetes complications in adults with type 2 diabetes were selected according to pre-specified criteria. Two investigators independently extracted data and evaluated all studies. If data were not reported in a comparable format, data were obtained from authors, presented as minimally adjusted ORs (and 95% CIs) per 1 year increase in age at diabetes diagnosis, adjusted for current age for each outcome of interest. The study protocol was recorded with PROSPERO International Prospective Register of Systematic Reviews (CRD42016043593). Results: Data from 26 observational studies comprising 1,325,493 individuals from 30 countries were included. Random-effects meta-analyses with inverse variance weighting were used to obtain the pooled ORs. Age at diabetes diagnosis was inversely associated with risk of all-cause mortality and macrovascular and microvascular disease (all p < 0.001). Each 1 year increase in age at diabetes diagnosis was associated with a 4%, 3% and 5% decreased risk of all-cause mortality, macrovascular disease and microvascular disease, respectively, adjusted for current age. The effects were consistent for the individual components of the composite outcomes (all p < 0.001). Conclusions/interpretation: Younger, rather than older, age at diabetes diagnosis was associated with higher risk of mortality and vascular disease. Early and sustained interventions to delay type 2 diabetes onset and improve blood glucose levels and cardiovascular risk profiles of those already diagnosed are essential to reduce morbidity and mortality.

Objective: Examine the health and economic impact of extending screening intervals in people with Type 2 diabetes (T2DM) and Type 1 diabetes (T1DM) without diabetes-related retinopathy (DR). Setting: Diabetic Eye Screening Wales (DESW). Study design: Retrospective observational study with cost-utility analysis (CUA) and Decremental Cost-Effectiveness Ratios (DCER) study. Intervention: Biennial screening versus usual care (annual screening). Inputs: Anonymised data from DESW were linked to primary care data for people with two prior screening events with no DR. Transition probabilities for progression to DR were estimated based on a subset of 26,812 and 1232 people with T2DM and T1DM, respectively. DCER above £20,000 per QALY was considered cost-effective. Results: The base case analysis DCER results of £71,243 and £23,446 per QALY for T2DM and T1DM respectively at a 3.5% discount rate and £56,822 and £14,221 respectively when discounted at 1.5%. Diabetes management represented by the mean HbA_1c was 7.5% for those with T2DM and 8.7% for T1DM. Sensitivity analysis: Extending screening to biennial based on HbA_1c, being the strongest predictor of progression of DR, at three levels of HbA_1c 6.5%, 8.0% and 9.5% lost one QALY saving the NHS £106,075; £58,653 and £31,626 respectively for T2DM and £94,696, £37,646 and £11,089 respectively for T1DM. In addition, extending screening to biennial based on the duration of diabetes > 6 years for T2DM per QALY lost, saving the NHS £54,106 and for 6-12 and > 12 years for T1DM saving £83,856, £23,446 and £13,340 respectively. Conclusions: Base case and sensitivity analyses indicate biennial screening to be cost-effective for T2DM irrespective of HbA_1c and duration of diabetes. However, the uncertainty around the DCER indicates that annual screening should be maintained for those with T1DM especially when the HbA_1c exceeds 80 mmol/mol (9.5%) and duration of diabetes is greater than 12 years.

In this paper, a method is presented for superimposition (i.e. registration) of eye fundus images from persons with diabetes screened over many years for diabetic retinopathy. The method is fully automatic and robust to camera changes and colour variations across the images both in space and time. All the stages of the process are designed for longitudinal analysis of cohort public health databases where retinal examinations are made at approximately yearly intervals. The method relies on a model correcting two radial distortions and an affine transformation between pairs of images which is robustly fitted on salient points. Each stage involves linear estimators followed by non-linear optimisation. The model of image warping is also invertible for fast computation. The method has been validated (1) on a simulated montage and (2) on public health databases with 69 patients with high quality images (271 pairs acquired mostly with different types of camera and 268 pairs acquired mostly with the same type of camera) with success rates of 92% and 98%, and five patients (20 pairs) with low quality images with a success rate of 100%. Compared to two state-of-the-art methods, ours gives better results.