Vascular Parameters of the Retina Show Accuracy in Evaluating Stroke Risk

Researchers have identified a retinal vascular “fingerprint” consisting of 29 indicators, which were shown to be accurate in evaluating stroke risk, according to findings of a UK Biobank study published in Heart. The retinal vascular fingerprint performed as accurately as traditional methods for determining stroke risk with the benefit of a noninvasive screening approach. The 29 parameters comprising this fingerprint fall under categories such as caliber, density, and complexity and arterial inflection count tortuosity, all of which are measured using fundus images of the retina.

The observational study evaluated fundus images of 45,161 individuals from the UK Biobank who went on to experience 749 strokes in total during a median follow-up period of 12.5 years. A machine learning technique was used to identify a total of 118 retinal vascular parameters, of which 29 were found to be significantly associated with risk of a first-time stroke.

• For the density parameters, which comprised over half of the fingerprint parameters, a standard deviation (SD) change was associated with a 9.8% to 19.0% increase in stroke risk.
• A SD change in caliber parameters was associated with a 10.1% to 14.1% increase in stroke risk.
• Each SD decrease in parameters of complexity and arterial tortuosity was associated with a 10.4% to 19.5% increase in stroke risk.

Additionally, the researchers used statistical analysis to compare the predictive power of the fingerprint with that of a traditional risk stratification model. The receiver operating characteristic (ROC) curve increased to .752 after incorporating the retinal vascular fingerprint, outperforming the traditional model (.739; P<.001). The authors noted that more research is necessary to confirm the results of this observational study.