Cutting-edge advance on the topic is OCT-angiography revealing preclinical changes in retina of persons with diabetes opening a new therapeutic window.

Good luck !

The overall objective of this work is to use information technology and individual risk assessment by mathematical algorithms to increase the efficacy of screening systems in medicine. The specific case is diabetic retinopathy screening, which has been successful in reducing diabetic blindness in many countries but may provide opportunity for improved cost-effectiveness.

Initially annual screening for diabetic retinopathy was the rule and is still recommended by WHO and many professional societies and practised in many countries, including the UK. However, more and more evidence has shown that annual screening is an overkill for low-risk patients with diabetes. Kristinsson et al, showed that diabetics without retinopathy may safely be screened every other year and this was confirmed by a 10 year experience. Biennial screening is practiced in many screening programmes for diabetic retinopathy and simple risk stratification has been explored.

More recently, we proposed to use individual multifactor risk assessment for diabetic retinopathy progression to control screening frequency and further economise screening programmes for diabetic retinopathy. Aspelund et al7 developed a mathematical algorithm that calculates individual risk for progression to sight-threatening diabetic retinopathy and can be used to control the frequency of diabetic eye-screening visits. The algorithm receives clinical data: type and duration of diabetes, HbA1c, systolic blood pressure, gender and the presence and grade of retinopathy. These data are used to calculate an individual's risk of developing sight-threatening retinopathy (STR), that is, proliferative diabetic retinopathy and/or clinically significant diabetic macular oedema. This algorithm has been validated successfully in Denmark, the Netherlands and Spain. It predicts about 80% of the risk of progression of diabetic retinopathy and recommends a screening interval according to individual risk. This allows reduction in mean screening frequency by over 50% (depending on interval ceiling) with corresponding reduction in costs. At the same time, high-risk patients with diabetes are screened more frequently, up to every 6 months depending on individual risk.

There is a slight difference in staging of diabetic retinopathy in the Danish, Dutch, Spanish and Icelandic cohorts initially used to calibrate the algorithm and the English screening system, which uses R0–3 with the subdivision of the third R3A (active proliferation) and R3AS (stable proliferation). M1 describes maculopathy in accordance with the English National Diabetic Eye Screening Program definition. (http://diabeticeye.screening.nhs.uk/getdata.php?id=11653). In order to use the algorithm in the English screening system, it must be calibrated for this staging and preferably in an English diabetic cohort.

The aim of this analysis is threefold: First to investigate whether the algorithm can be used to predict the risk of R2 (preproliferative retinopathy), R3A (active proliferative retinopathy) and M1 (diabetic maculopathy). We also want to estimate the possible reduction in screening frequency in this cohort, while maintaining safety standards. The third goal is to establish the risk profile for the cohort to recognise how many are at low risk and how many have high risk for progression within 2 years.

This study is based on data of 9690 individuals with diabetes in England who are participating in the English National Diabetes Eye Screening Programme. Data was made available on research consented patients, and their anonymised data included their retinopathy stage in the R and M scales annually for 3 years. Clinical data was also available on type and duration of diabetes, HbA1c, systolic blood pressure, gender and the presence and grade of retinopathy. Three individuals with systolic blood pressure less than 80 mm Hg were excluded from the analysis. Based on the algorithm of Aspelund et al, the subjects’ risk of developing R2, R3A or M1 was calculated and screening intervals were recommended. The calculations were done using clinical data gathered in the year 2010 and compared with clinical outcome in the year 2012.

The algorithm was originally designed to estimate the risk of developing sight-threatening retinopathy (STR, either diabetic macular edema (DME) or profilerative diabetic retinopathy (PDR)). In order to predict the risk of the occurrence of retinopathy stage R2, R3A or M1, instead of STR), a calibration coefficient was calculated for each retinopathy stage, and gender and type of diabetes. The coefficient was the observed proportion of each outcome divided by the average risk estimate for STR. An estimate of the risk of developing R2, R3A or M1 was calculated by multiplying the risk estimate of the algorithm with the corresponding calibration coefficient.

The recommended screening intervals were calculated so that the average cumulative risk of developing R2, R3A or M1 within the screening period was kept identical to the cumulative incidences of the corresponding retinopathy grading that were observed with annual screening.

A receiver operating characteristic (ROC) curve is a plot of the true positive rate (sensitivity) against the false positive rate (1−specificity) for the different decision thresholds of a diagnostic test. The area under the ROC curve is an estimate of the capacity of the diagnostic test to distinguish individuals with versus those without disease. It can range from 0.5 to 1, where an area of 1 represents a perfect test and an area of 0.5 represents a worthless test. ROC curves were plotted for the ability of the algorithm to predict development of R2, R3A and M1 within 2 years for the population in total as well as patients with type I and type II diabetes separately. ROC curves for the ability of any retinopathy grading were also drawn for patients with type I and type II diabetes separately.

Thank you Marianne and Entedhar for these timely and insightful responses.

Entedhar is there is any online link to the mathematical algorithm used in this study?

Thanks,

Nyawira

Control all of the usual athrothrombotic disease risk factors: cigarette smoking, dyslipidemia, hypertension , and of course the blood sugar levels. If this is done, diabetic retinopathy is rare.