Educational Levels and Cardiovascular Risk Scores to predict Deep White Matter Hyperintensities in Fazekas Scale

Abstract Background Cardiovascular risk factors (CVR) are associated with an increased risk in developing dementia through lifespan. Additionally, magnetic resonance imaging (MRI) can assess the cerebrovascular burden and atrophy patterns that are essential for dementia's differential diagnosis. Objective To conduct an analysis of the relationship of sociodemographic factors, white matter burdens with FAZEKAS visual scale on brain MRI, and CVR estimates. Methods This is a cross‐sectional study, using data from the Cog‐Aging cohort study from 2018 to 2024. Forty‐one participants were recruited. MRI was analysed by two experienced and independent physicians to determine Fazekas deep white matter (DWM) and periventricular white matter (PVWM) T2 hyperintensities scores. The Cardiovascular Risk Score (QRISK3) and the Atherosclerotic Cardiovascular Disease (ASCVD) were selected to estimate CVR. Differences in Fazekas scores between CVR, diagnosis, and sociodemographic groups were explored with Mann‐Whitney and Kruskal‐Wallis analysis. Significant associations were further explored in Logistic Regression adjusted for education to determine the probability of the calculator results predicting changes in the FAZEKAS. Skin color, age and gender were removed from the regression because they are variables used in both calculators to assess risk. This study was approved by the UFMG Ethics Committee. Results The mean age was 76.65 years (SD= 7.36) and the median years of education was 7 (IQR= 8.00). Differences in Fazekas scores were observed only for DWM between educational level groups. Logistic regression of CVR, showed that the QRisk (OR=1.067, p = .027) was slightly better at predicting DWM score of 2 or more, than ASCVD (OR=1.053, p = .023). When adjusted for education, both CVR scores lost significance, but the QRisk model was slightly better (AUC= .809, Nagelkerke R2= .361 / vs .787 and .329). Logistic Regression model with only educational level was significant (OR= .810 p.= .008) and when adjusted with CRV scores, it maintained significance (OR= .823, p = .017 QRisk; OR= .844, p = .036 ASCVD). Conclusion These findings suggest that the educational level variable was better to predict a worse DWM (≥2) score than CVR scores. These are preliminary results and more studies with a higher sample with cardiovascular risk scores and MRI are needed.