Retinal microvascular features are associated with CMR measures of subclinical cardiovascular dysfunction

Background Microvascular dysfunction is a key component of many cardiovascular (CV) diseases. Assessing the retinal microvasculature through retinal imaging may therefore provide a window into evaluating a range of CV diseases. This study sought to generate hypotheses regarding relationships between different retinal microvascular features (RVF) and measures of subclinical CV dysfunction derived from cardiovascular magnetic resonance (CMR) imaging. Methods 182 participants with type 2 diabetes enrolled in the UK Imaging Diabetes Study (UKIDS) with CMR image data were considered for inclusion in this cross-sectional study. Fifteen CMR measures of cardiac structure, function, tissue characterisation, adiposity, and aortic distensibility were derived. One-hundred-twenty-eight participants (70%) were found to have eligible retinal images. An artificial intelligence (AI)-enabled retinal imaging analysis tool (QUARTZ) quantified eight RVFs from each participant's retinal image: arteriolar and venular diameter, area, calibre uniformity, and tortuosity. Correlation analysis shortlisted RVF-CMR variable pairs for multivariable regression. Regression coefficients represent change per 1 standard deviation (SD) increase in RVF. Results Sixteen RVF-CMR regression pairs were shortlisted for regression, and five remained associated after adjustment for potential confounders. Per 1-SD increase in venular tortuosity was associated with a 0.5ms greater left ventricular (LV) T2 mean, 0.6% worsening in LV global longitudinal strain, and a 2 mL greater left atrial max volume. Per 1-SD increase in arteriolar calibre uniformity and retinal venular area were associated with 9ms lower LV T1 mean and 0.2x10-3mmHg-1 greater proximal descending aortic distensibility respectively. No significant associations were found between RVF and LV volumetric or functional measures, or adiposity. Conclusions In a diabetic cohort, we identified novel and biologically plausible associations between RVF and CMR measures of subclinical CV dysfunction. This provides new insight into the relationship between the retinal and systemic vascular beds and supports the potential role of retinal imaging in evaluating CV dysfunction prior to onset of overt disease.