Date of Award

January 2024

Document Type


Degree Name

Medical Doctor (MD)



First Advisor

Mathieu F. Bakhoum


Matesva Mitchelle1, Mathieu F. Bakhoum11Department of Ophthalmology and Visual Sciences, Yale University School of Medicine, New Haven, CT

Background and Purpose: Diabetic Retinopathy (DR) is a leading cause of vision loss worldwide. Its impact is magnified due to the widespread prevalence of diabetes, affecting an estimated 463 million people globally. Traditional screening methods, like fundus examination, fall short in efficiency and accessibility for large numbers. Furthermore, a lack of access to specialized ophthalmological care due to geographic and socioeconomic barriers exacerbates the problem. Considering these challenges, we sought to investigate the utility of more accessible, convenient DR diagnostic tools through two aims: 1) a urine-based test to assess levels of Vascular Endothelial Growth Factor (VEGF), a key mediator of DR pathogenesis and 2) optimizing droplet digital PCR (ddPCR) for amplification of long terminal repeats (LTRs) to quantify cell free DNA (cfDNA) in urine as a diagnostic tool for diabetic retinopathy.

Study Design and Methods: For the evaluation of the assay for urinary VEGF, participants who were evaluated in the retina clinic were prospectively enrolled, after obtaining informed consent. Inclusion criteria were: a diagnosis of type 2 diabetes mellitus with or without DR, and controls without diabetes mellitus. Exclusion criteria included administration of anti-VEGF injections within the preceding one month. Urine samples were centrifuged at 2,000 rpm for 20 minutes and stored at -80 0C. VEGF-A levels were quantified using AlphaLISA, a rapid and highly sensitive bead-based immunoassay with a wide dynamic range of detection (2.2–100,000 pg/mL). Levels of urine albumin, creatinine, 1and osmolality were also measured using calorimetric tests. To account for the differences in urine tonicity, urinary VEGF concentrations and urine albumin levels were divided by urine creatinine to obtain normalized values. For the cfDNA quantification, ddPCR conditions (primer concentration, annealing temperature, PCR cycles, elongation time and positive control concentration) were optimized to effectively quantify cfDNA in urine.

Main Outcome Measures:For the evaluation of urinary VEGF, the main outcome measure was a diagnosis of DR. We evaluated the association between normalized urinary VEGF concentration and the presence of DR. For the cfDNA quantification, descriptive statistics, 1D plots and Poisson’s uncertainty curves at 95% confidence level were obtained using the Crystal Miner software (Stilla Technologies), following the manufacturer’s recommendations. Target concentrations in copy number per µl reaction were then automatically calculated by the Crystal Miner software.

Results:For the evaluation of urinary VEGF, there were 30 participants who met the inclusion criteria. These included 4 non-diabetic individuals, 10 diabetic individuals without retinopathy, 9 with non-proliferative DR, and 7 with proliferative DR. Normalized urinary VEGF concentrations did not differ significantly across participants with proliferative DR (0.34 ± 0.32), non-proliferative DR (0.29 ± 0.26), and those without DR (0.44 ± 0.45) (P>0.05 for all). A multivariate logistic regression analysis revealed that after adjusting for age, sex, calculated eGFR and normalized urine albumin, normalized urinary VEGF was not a significant predictor of the presence of DR. In addition, a multivariate logistic regression analysis revealed that after adjusting for age, sex, calculated eGFR, urine osmolality and normalized urine albumin, raw urinary VEGF was not a significant predictor of the presence of DR. For the cfDNA quantification, we optimized ddPCR parameters for analyzing cfDNA in urine. Our findings identified 64.3°C as the optimal annealing temperature, a 1:1000 dilution of extracted DNA from human cells, and 28 PCR cycles as the most effective conditions. Despite subsequent attempts to address oversaturation and unclear differentiation between positive and negative droplet populations, such as using freshly extracted DNA and adjusting elongation time and primer concentrations, we encountered challenges with poor band separation. Running ddPCR on undiluted urine samples revealed widespread oversaturation, hindering our ability to establish a clear link between cfDNA levels and DR status.

Conclusion: We found no significant association between urinary VEGF concentrations and presence of DR, even after adjusting for potential confounders. For the second aim on cfDNA quantification, we optimized ddPCR for quantifying cfDNA from urine samples, providing insights into assay performance factors. Our study highlights the challenges associated with utilizing ddPCR for LTR amplification in urine. While ddPCR effectively quantifies minute amounts of cfDNA, further optimizations are necessary to achieve clear distinction between positive and negative populations. Further optimization, particularly in diluting urine samples, is crucial to address oversaturation issues during cfDNA quantification via ddPCR, preceding the exploration of the correlation between urinary cfDNA and diabetic retinopathy (DR).


This thesis is restricted to Yale network users only. It will be made publicly available on 04/30/2025