Does inclusion of cognitive measures improve the predictive performance of SCORE2 for future cardiovascular events in older adults?

Abstract Funding Acknowledgements Type of funding sources: Public grant(s) – EU funding. Main funding source(s): National Institute for Research of Metabolic and Cardiovascular Diseases (Programme EXCELES, ID Project No. LX22NPO5104) - Funded by the European Union – Next Generation EU. Introduction The cardiovascular disease (CVD) risk prediction model SCORE2 uses several biomedical and behavioural factors. It is likely that other factors also contribute to CVD risk prediction. Cognitive impairment and CVD are associated and share several risk factors. While the causality of this relationship remains unclear, adding cognitive measures to risk prediction may improve the prediction of future CVD events. Objective We tested the hypothesis that including simple measures of cognitive functions improve the prediction of SCORE2 for future CVD events among middle aged and older subjects. Methods We used data on 13,391 Polish, Lithuanian and Czech adults without CVD at baseline (45–69 years) from the Health, Alcohol and Psychosocial factors In Eastern Europe cohort. Incident cases of myocardial infarction (MI), stroke, CVD mortality and composite (fatal and non-fatal) CVD events were identified over a 10-year follow-up. Baseline cognitive measures included immediate and delayed word recall (verbal memory and learning), animal naming (verbal fluency) and letter cancellation (attention, mental speed and concentration). Using competing-risks regression, relationships between cognitive functions and CVD outcomes were examined. Improvement in Area Under Curve Receiver Operator Characteristic (AUC-ROC) was used to compare predictive performance for CVD events of models including SCORE2 variables combined with cognitive tests vs SCORE2 alone. Results The incidence rates of MI, stroke, CVD mortality and composite CVD events were 5%, 4%, 4% and 11%, respectively. In multivariable analyses, all four cognitive measures were inversely associated with CVD outcomes; e.g. one standard deviation increase in immediate word recall score was associated with a 28% and 17% reduction in risk of CVD mortality and composite CVD events, respectively (all p-values <0.001). Adding cognitive measures to SCORE2 variables slightly improved the AUC-ROC for prediction of MI, CVD mortality and composite CVD events. For example, for the composite CVD measure, inclusion of all four cognitive measures increased the AUC-ROC from 0.6948 to 0.7030 (p<0.001) and similar improvement was seen for word recall. The results were consistent across countries. Conclusion Inclusion of a simple cognitive test in the assessment of cardiovascular health is feasible and it may improve prediction of future CVD risk, although in our data the improvement was not clinically important (despite being statistically significant). Further studies are needed to assess this proposition in larger prospective datasets.