Retinal Microvascular Analysis Predicts Cardiovascular Outcomes in Diabetic Patients

Iffy Mordi, Colin NA Palmer, Stephen Hogg, Ruixuan Wang, Tsz Li, Aaron Tee, Lucia Ballerini, Emanuele Trucco, Thomas Macgillivray, Gareth McKay, C. Lang, Alexander S F Doney

Research output: Contribution to conferencePaperpeer-review


Background: Microvascular dysfunction contributes to the development and progression of cardio-metabolic disease and is associated with increased cardiovascular (CV) mortality. However assessing microvascular health is difficult in clinical practice. The retina provides an observable microvascular bed that might offer important insights into global vascular health and patients with diabetes are offered regular retinal screening to manage risk of diabetic retinopathy (DR). The VAMPIRE (Vessel Assessment and Measurement Platform for Images of the REtina) software tool enables high-throughput, semiautomatic measurement of retinal vascular parameters (RVP) including vessel diameter, tortuosity and fractal dimension (Df – a marker of vessel complexity). We hypothesised that RVPs extracted from standard diabetic retinal screening photographs would inform on global cardiovascular risk in patients with diabetes. Methods: We evaluated 4,467 patients with diabetes in the Genetics of Diabetes Audit and Research in Tayside Scotland (GoDARTS) study in whom VAMPIRE had been used to extract RVPs from standard diabetic retinal screening photographs. We exploited advanced electronic medical records capabilities in GoDARTS to link this information to cardiovascular outcomes and echocardiographic parameters. We used logistic regression to investigate the association between RVPs and echocardiography parameters and Cox’s regression to investigate their association with major adverse cardiovascular events (MACE - cardiovascular death, hospitalisation for myocardial infarction or stroke).ResultsMedian follow-up was 6.6 years from date of retinal photograph from which vascular measures were taken. Mean age at entry was 71±12 years and 57% of the cohort were male. After adjustment for relevant clinical variables including age, gender, duration of diabetes, mean HbA1c and systolic blood pressure prior to study entry, prior MI or stroke and insulin use, both arterial and venular tortuosity were associated with echocardiographic left ventricular hypertrophy (LVH) (OR 1.02, p=0.01 and OR 1.04, p=0.000004 respectively). Df was negatively associated with both LVH and left ventricular systolic dysfunction (LVSD) (OR 0.81, p=0.007 and 0.82, p=0.006 respectively). Neither retinal arterial or venular diameter were significantly associated with either LVSD or LVH on echo. MACE occurred in 614 patients following the date of the measured retina. After adjustment for these same clinical variables, increased venular tortuosity was associated with adverse outcome (HR 1.22, p= 0.00005). Conversely, increased Df was associated with improved outcome (HR 0.34, p=0.027). Tortuosity and fractal dimensions were independent, therefore these were summed to make a score. Patients in the highest quartile of combined score were significantly more likely to suffer an adverse event (HR 2.38, 95% CI 1.54-3.70, p=0.0001).Conclusion: While increased venous tortuosity was associated with adverse outcome, increased fractal complexity of the vessels (increased branching) appeared to be protective. The latter may be an indication of increased ability to maintain a collateral supply in the presence of ischaemia. Diabetes retinal screening images available through national programs may provide important independent risk indicators for global CVD risk in patients with diabetes that go beyond conventional screening for DR. RVPs may contribute to further insights into pathophysiology of CVD.Figure. Adjusted Cox Regression for the association of quartiles of venular tortuosity (left) and Fractal Dimension (right) with the Primary Outcome: Kaplan-Meier curve of combined retinal score and the primary outcome. Log-rank p<0.0001

Fingerprint Dive into the research topics of 'Retinal Microvascular Analysis Predicts Cardiovascular Outcomes in Diabetic Patients'. Together they form a unique fingerprint.

Cite this