Endocrine Abstracts

Fixed annual interval screening for referable diabetic retinopathy (RDR) irrespective of the calculable risk of it developing is wasteful and unsustainable in the era of Precision medicine. Prediction models exist that can identify the at-risk and low-risk subgroups and assign screening intervals based on that risk. Using data on 2770 participants followed for a mean of 3.5 years of annual retinopathy grading results combined with twice yearly systemic risk factor measurements; we identified and published the biomarkers most correlated with progression to referable retinopathy status in an Irish population for the first time. The strongest association was PRIOR RETINOPATHY SEVERITY (HR 4.02) followed by most recently recorded SYSTOLIC BP (HR 1.29)/HbA1c (HR 1.22) SERUM TRIGLYCERIDES (HR 1.10). Furthermore, we carried out the first external validation of prediction models developed using similar data assembled in England. Its accuracy of prediction was evaluated by comparing predicted with observed outcomes. Its predictive accuracy was compared with the commercially available RETINARISK app which empowers users with the means of tracking their own disease progress. Whilst the recently introduced Irish screening algorithm which assigns biennial screening intervals using retinopathy screening results alone; our analysis demonstrates that precision of prediction and hence risk stratification are enhanced by inputting values on a limited series of systemic risk factors relevant to the population of study. A ten-fold difference in risk stratification between lowest and highest risk groups was found in the English models. We demonstrate that there would be a >70% probability that a randomly selected subject from the screening cohort who did in fact develop RDR would have been allocated to the higher risk score category by the models tested. This means that their next scheduled screening appointment would not be extended from 1 to 2 years based on advice from the prediction model.