VIENNA – A risk model based on a single assessment of hemoglobin A1c and other parameters was able to differentiate well between patients who had a low and high risk for progression of sight-threatening diabetic retinopathy, researchers reported at the annual meeting of the European Association for the Study of Diabetes.
Although further validation is needed, the risk model “would be suitable for personalized screening intervals,” said presenting author Dr. Irene Stratton of the Gloucestershire (England) Hospitals NHS Foundation Trust. In England, the current recommendation is to screen everyone with diabetes annually using digital retinal photography.
Because of the increasing numbers of patients that require annual screening, health budgets for diabetic eye screening are being stretched and alternatives are desirable. In England, such screening considers both retinopathy and macropathology, Dr. Stratton said.
The risk model was derived and tested based on data from 14,000 patients with no or mild nonproliferative retinopathy treated in the Gloucestershire area. The investigators looked at HbA1c levels in the 12 months prior to diabetic eye screening to determine if they could predict which patients did and did not develop sight-threatening retinopathy. The risk model also considered the baseline retinopathy status in both eyes of patients; systolic and diastolic blood pressure; measures of kidney function and lipids; the time to develop retinopathy from diagnosis; and the type of diabetes.
“We found that the most important piece of information that went into the model was the grading at the baseline screening episode,” Dr. Stratton said. “So patients who had no retinopathy in either eye were at lowest risk, patients with background retinopathy in one eye had about a doubling of risk, and patients with background retinopathy in both eyes were at a much higher level of risk.”
The next most important parameters were the time since the first mention of a diagnosis of diabetes and HbA1c in the year prior to screening. Total cholesterol in the year prior to screening also was a factor for consideration.
Hazard ratios for the development of diabetic retinopathy were 7.13 for patients with mild retinopathy in both eyes at baseline and 2.56 for those with mild retinopathy in one eye. The hazard ratios increased by 1.28 for every 10-mmol/mol increase in HbA1c in the past 12 months, by 1.20 for every 5-year increase in the duration of diabetes, and by 1.12 for every 1-mmol/L increase in total serum cholesterol.
Three study populations, which altogether comprised almost 20,000 patients with diabetes, were used to validate the model; the largest population included more than 17,000 individuals. The main difference between the screening programs was the ethnic mix, Dr. Stratton highlighted. White patients dominated in the largest screening cohort, at 98%, but to a lesser extent in the other cohorts, at 47% (of 1,223 patients) and 81% (of 1,083). About half the patients in the cohorts were women, and the duration of diabetes ranged from 2.9 to 4.5 years. The majority (95%) of patients screened had type 1 diabetes. HbA1c ranged from 6.3% to 8.2% overall.
The investigators stratified patients according to quintiles of risk using the model, and found that the quintiles correlated very well with the chances of patients developing sight-threatening eye disease in each of the three validation cohorts. Comparing the lowest- with the highest-risk quintiles, the rate of progression to sight-threatening retinopathy was 1-3 and 55-79 per 1,000 per patient-years, respectively. The overall event rate was around 20 per 1,000 patient years.
“Further validation in other screening programs and ethnic groups is required,” Dr. Stratton concluded.
The study was funded by a grant from the U.K. National Institute for Health Research, Health Technology Assessment Programme; and the Gloucestershire Hospitals National Health Service Foundation Trust. Dr. Stratton had no conflicts of interest.