Abstract
Purpose :
Diabetic Retinopathy (DR) is the leading cause of blindness in 20-74 year-old U.S. adults. Genetic factors, including single nucleotide polymorphisms (SNPs), have been correlated with DR susceptibility and progression. Our goal is to develop a platform to explore novel associations between DR phenotypes and genetic variants.
Methods :
We queried the Synthetic Derivative (SD), a de-identified database from VUMC’s electronic health record linked to DNA samples. A combination of ICD and CPT codes was used to classify all SD records by presence of diabetes mellitus (DM), DR, non-proliferative and proliferative DR (NPDR, PDR), and diabetic macular edema (DME). We used Quanto software to calculate the expected statistical power of this platform to detect associations between vision threatening DR (VTDR) and SNPs of varying allele frequencies (AF). To confirm the accuracy of our code-based search criteria, presence of DM, DR, NPDR, PDR and DME was confirmed by manual review for a random sub-sample. Data regarding additional characteristics (risk factors, comorbidities, diagnostics, therapeutics) were also manually collected. We used Pearson’s analysis to correlate the prevalence of these characteristics with DR susceptibility and progression (control vs DR, NPDR vs PDR, and no DME vs DME) in our sub-sample.
Results :
A total of 5447 controls, 1168 NPDR, 734 PDR, and 441 DME cases met ICD/CPT code-based search criteria. This platform has >80% power to detect odds ratios >1.4 for SNPs with AF >35% using pre-existing genetic data, and for SNPs with AF >10% among all records. DR status and clinical characteristics were confirmed by manual review of 155 controls, 134 NPDR, 178 PDR, and 188 DME cases. Among these, Type 1 DM, nephropathy, and neuropathy were more prevalent in DR cases than controls (p<0.00001) and in PDR than NPDR cohorts (p<0.05). Type 2 DM, obesity, and hypertension were more prevalent in cohorts with DME than without (p<0.05) but had no evidence of association with progression to PDR. There was no evidence of association between gender, smoking, mortality, and hyperlipidemia with DR susceptibility or progression.
Conclusions :
Our platform is clinically relevant and adequately powered to detect associations between SNPs and VTDR. We will use this resource to explore novel associations that may provide insights into DR pathophysiology, diagnosis, and management.
This is a 2021 ARVO Annual Meeting abstract.