Abstract Body: Introduction: Stroke can contribute to age-related cognitive decline, directly affecting the quality of life and post-stroke functioning. While cognitive impairment is usually associated with chronological age, the association between biological/electrocardiographic age (ECG-age) and cognition is unclear. We used a machine learning model to predict the ECG-age of participants directly from their ECG waveform data in the UK Biobank. We determined the association between participants’ ECG-age and their scores on various digital cognitive tests.
Methods: In the UK Biobank, about 500,000 participants aged between 40 and 69 were recruited from across the United Kingdom. Electrocardiogram data was collected using a 12-lead ECG at rest. Deep neural network (DNN) was used to predict the ECG-age of participants from their raw ECG waveform data. In our main analysis, participants with available ECG data and cognitive test scores were selected (n=63800). Δage was defined as the difference between ECG-age and reported age. We used a multivariable linear regression model to evaluate the association between Δage and digital cognitive tests adjusted for reported age, sex, and education. Additionally, we defined aging according to these 3 categories: decelerated aging: participants with Δage below -5, normal aging: participants with an absolute value of Δage below mean absolute error, and accelerated aging: participants with Δage above 5.
Results: Compared to the normal aging group, the accelerated aging group (Δage>5) showed a significant decline in cognitive performance, whereas the decelerated aging group (Δage<-5) performed better on various cognitive assessments. The accelerated aging group showed negative β values in 6 out of the 8 cognitive tests analyzed, implying a negative change in performance in those tests compared to the normal aging group. On the other hand, the decelerated aging group outperformed the normal aging group in 6 out of 8 tests, showing positive β values in those tests. β and SE values describe the change in effect sizes and standard error following a 10-year change in the Δage. The risk of cognitive decline was higher in the accelerated aging group (p<0.016).
Conclusion: In a large population-based cohort, a strong association was observed between ECG-age and cognitive performance. Given the wide availability of ECG in cardiovascular diagnosis, including stroke, accelerated ECG aging could serve as a risk indicator for cognitive impairment.