Study population
The UK Biobank is a large-scale prospective population-based cohort study of 502,656 UK residents aged 40-69 who were registered with the National Health Service (NHS). Overall study protocols and data are available elsewhere [24]. Briefly, baseline assessments were carried out between 2006 and 2010 at 22 assessment centers across the UK. Participants completed electronic questionnaires to provide information on socio-demographics, lifestyle, environmental exposures, medical history, and cognitive functions. Physical examinations including blood pressure, heart rate, grip strength, anthropometry and spirometry were performed for all participants. Biological samples, including stored blood, urine and saliva samples, were taken. The follow-up of medical conditions was carried out mainly by means of data linkages with hospital records and mortality registers.
This study has been reviewed and approved by the National Information Governance Board for Health and Social Care and the NHS North West Multicenter Research Ethics Committee (11/NW/0382) and the Biobank Consortium (Application No. 62489). Because we used anonymized data in a public dataset, the Guangdong Provincial People’s Hospital Medical Research Ethics Committee waived the requirements for obtaining ethics approval. The study was carried out in accordance with the Declaration of Helsinki. All participants gave their informed consent.
Fundus photography
Between 2009 and 2010 eye examinations were introduced at six assessment centers across the UK [25]. Non-mydriatic retinal 45° fundus and optical coherence tomography (OCT) imaging of the optic disc and macula were captured using one spectral domain OCT for each eye (Topcon 3D OCT 1000 Mk2, Topcon Corp, Tokyo, Japan). Initially, ophthalmic examinations were performed on 66,500 participants, resulting in a total of 131,238 fundus images.
Deep learning model for age prediction
A total of 80,169 images from 46,969 participants passed the image quality check and were included in the analysis. Participant characteristics stratified by the number of images passing quality control have been described in detail in Supplementary File 1: Table S1. Among 46,969 participants, 11,052 participants reported no prior illness at baseline. The DL model for age prediction was constructed based on fundus images of disease-free participants. To maximize available data, binocular images, if available, were used for training and validation. The association between retinal age difference and stroke was investigated using images of the remaining 35,304 participants who had no history of stroke at baseline. Right eye images were included in the test set to predict retinal age and left eye images were used only if right eye images were not available.
Methods for predicting retinal age using DL models have been described in detail in a previous study [26]. Our previous study evaluated the performance of the DL model for age prediction. The DL model accurately predicted retinal age, as evidenced by a strong correlation of 0.80 (P<0.001) between predicted retinal age and chronological age, as well as an overall mean absolute error (MAE) of 3.55 years. Attention maps extracted from the DL model for age prediction mainly highlighted areas around retinal vessels in fundus images.
Definition of Retinal Age Gap
The retinal age gap was defined as the difference between the retinal age predicted by the DL model based on the fundus images and the chronological age. A positive retinal age gap indicated an “older” looking retina, while a negative gap indicated a “younger” looking retina.
Stroke determination
Stroke was determined by the UK Biobank Outcome Adjudication Group and was defined by codes 430.X, 431.X, 433.X1, 434.X1, 436.X in the 9th edition of the International Classification of Diseases ( ICD-9) and ICD-10 codes I60, I61, I63 and I64. Stroke events were derived from linked electronic health records, including hospital admissions and diagnosis records from hospitals in England, Scotland and Wales, as well as cause of death obtained from national death registers. The date of the first known stroke after the initial assessment was recorded. The follow-up period for each participant was defined from the date of enrollment in the UK Biobank study until February 28, 2018 (the last date of follow-up), or until the date of the first known stroke, depending on the first possibility.
Covariates
Covariates in the present analyzes included sociodemographic factors (initial age, gender, ethnicity, Townsend deprivation indices [TDI], level of education), lifestyle factors (smoking, alcohol consumption, level of physical activity) and general health. Baseline age and sex were obtained from central registry or self-reported questionnaires. Self-reported ethnicity was split into white or non-white. The TDI was an indirect measure of socioeconomic status based on postal code. Educational attainment was categorized as college/university or higher, or other. Smoking and alcohol consumption were categorized as current/previous users, or never. Physical activity level was categorized by meeting or not meeting the moderate/vigorous/walking recommendation. General health status was rated as excellent/good or fair/poor. Body mass index (BMI) was calculated as an individual’s weight in kilograms divided by their height in meters squared. Obesity was defined as a BMI of 30 kg/m2 or above. Diabetes mellitus was defined as any record of self-reported or physician-diagnosed diabetes mellitus, or the use of antihyperglycemic drugs or insulin. Hypertension was defined as self-reported or physician-diagnosed hypertension, or use of antihypertensive medications, or mean systolic blood pressure ≥ 130 mmHg or mean diastolic blood pressure ≥ 80 mmHg.
statistical analyzes
Continuous variables were reported as means and standard deviations (SD), while categorical variables were reported as numbers and percentages. Unpaired t-tests and chi-square tests were performed to examine differences in continuous and categorical variables, respectively. The log-rank test was used to compare survival distributions between different retinal age difference groups. Cox proportional hazards regression models were used to estimate the effect of retinal age difference on stroke risk. Each variable was assessed for the proportional hazards assumption and all met the assumption. Retinal age difference was entered into the models as a continuous variable (by one-year increment) and categorical variable (quintiles), respectively. Model I adjusted for age, gender, and baseline ethnicity. Model II was adjusted for all variables in Model I, plus TDI, education level, smoking status, alcohol status, physical activity level, diabetes mellitus, hypertension , obesity and general health. Logistic regression models were used to estimate the predictive value of the well-established classical model based on risk factors (including age, sex, smoking status, history of diabetes, systolic blood pressure and ratio total cholesterol/HDL cholesterol) [27] and the retinal age-based model of 10-year stroke risk. The area under the receiver operating characteristic curve (AUC) was used to describe pattern discrimination in predicting 10-year stroke risk.
A sensitivity analysis was performed to adjust the squared term of age in the final models in addition to age. We also investigated whether the retinal age acceleration residual (defined as the residuals from the regression of predicted retinal age against chronological age) was a biomarker of stroke in the second sensitivity analysis.
A double sided p a value < 0.05 indicated statistical significance. All analyzes were performed with R (version 3.3.0, R Foundation for Statistical Computing, www.R-project.org, Vienna, Austria) and Stata (version 13, StataCorp, TX, USA).
#Retinal #age #difference #predictive #biomarker #stroke #risk #BMC #Medicine