Diabetic retinopathy: towards optimizing prediction of treatment outcome

Diabetic eye disease is a common complication of diabetes, which can lead to vision loss. Two types of diabetic eye disease, diabetic macular edema (DME) and proliferative diabetic retinopathy (PDR), can cause severe vision loss. However, treatment outcomes for both diseases are not yet promising, and results can vary widely from patient to patient. Therefore, it is important to predict the outcome to optimize treatment effects for individual patients.
This thesis explores two novel biomarkers, hyperreflective dots (HRDs) on optical coherence tomography (OCT) and skin autofluorescence (SAF) measured with the AGE Reader, which have the potential to predict treatment outcomes in DME and PDR, respectively. Previous studies evaluating the predictive value of HRDs have yielded contradictory results, highlighting the need for standardized definitions and calculation methods. In this thesis, we applied standardized definitions. We developed a deep learning algorithm that enables a standardized quantification of HRD numbers on OCT in DME eyes. HRD numbers measured with this algorithm can partially predict the one-year treatment outcomes in DME eyes given the standardized treatment.
SAF can provide an estimate of people’s past glucose levels over a long period of time. High SAF values within the diabetes population may indicate chronic suboptimal blood glucose control and an increased risk of severe end organ damage, including the severe stage of PDR. However, SAF did not predict the visual outcomes in PDR patients requiring vitrectomy surgery.
This thesis extends our insight into the role of systemic and ocular biomarkers in diabetic eye disease and may help optimize individualized strategies to improve visual prognosis. However, further investigations are needed to better understand the clinical implications of both biomarkers.