Capturing population differences in rates of vascular aging using a deep learning electrocardiogram algorithm: a cross-sectional study

Background: Cardiovascular event rates increase with age in all populations. This is thought to be the result of multiple underlying molecular and cellular processes that lead to cumulative vascular damage. Apart from arterial stiffness based on pulse wave velocity there are few other non-invasive measures of this process of vascular aging. We have developed a potential biomarker of vascular aging using deep-learning to predict age from a standard 12-lead electrocardiogram (ECG). The difference between ECG predicted and chronological age ({delta}-age) can be interpreted as a measure of vascular aging. Methods: We use data collected in two cross-sectional studies of adults aged 40-69 years in Norway and Russia to test the hypothesis that mean levels of {delta}-age, derived from a deep-learning model trained on a US population, correspond to the known large differences in cardiovascular mortality between the two countries. Findings: Substantial differences were found in mean {delta}-age between populations: Russia-USA (+5{middle dot}2 years; 0{middle dot}7, 10 IQR) and Norway-USA (-2{middle dot}6 years; -7, 2 IQR). These differences were only marginally explained when accounting for differences in established cardiovascular disease risk factors. Interpretation: {delta}-age may be an important biomarker of fundamental differences in cardiovascular disease risk between populations as well as between individuals.

damage. Apart from arterial stiffness based on pulse wave velocity there are few other non-48 invasive measures of this process of vascular aging. We have developed a potential biomarker of 49 vascular aging using deep-learning to predict age from a standard 12-lead electrocardiogram 50 (ECG). The difference between ECG predicted and chronological age (δ-age) can be interpreted 51 as a measure of vascular aging.

52
Methods: We use data collected in two cross-sectional studies of adults aged 40-69 years in

53
Norway and Russia to test the hypothesis that mean levels of δ-age, derived from a deep-54 learning model trained on a US population, correspond to the known large differences in 55 cardiovascular mortality between the two countries.

78
We have developed a novel potential biomarker of vascular aging. This arose from work to 79 predict age from a deep learning analysis of 12-lead ECGs. 7 It was noted that while predicted 80 ECG-age was highly correlated with chronological age there were deviations between the two. 7

81
This led to the proposal that the direction and magnitude of the difference between ECG-age and 82 chronological age (delta-age (δ-age)) may be a measure of relative vascular aging: if δ-age is 83 positive, it would suggest a higher cumulative vascular damage and therefore, higher rate of 84 vascular aging relative to the reference population from which the age algorithm was derived.

85
This has been supported by our subsequent work that has shown δ-age to be prospectively 86 associated with mortality, 8 cardiovascular events 9 and cross-sectionally with a range of 87 established CVD risk factors including smoking and blood pressure. 10 There is also a strong 88 positive association between δ-age and pulse-wave velocity, 10 an established marker of vascular 89 stiffness, which has been regarded as one of the few direct markers of vascular aging.

90
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint  group. Given these differences in CVD mortality we hypothesised that the mean difference 106 between ECG predicted and chronological age (δ-age) for Russia would be positive, while the 107 equivalent mean value of δ-aged for Norway would be negative. We also investigated how far any 108 differences in mean delta-age between the Russian and Norwegian study populations could be 109 explained by differences in known CVD risk factors.  (Table S1). Mean δ-age was +5·2 years (0·7, 10 IQR) for the KYH (Russia) study 115 population, (Figure 1b) and -2·6 (-7, 2 IQR) for T7 (Norway) study population (Figure 1c). The

121
The loess smoothed fitted line for KYH ( Figure 1b) was above the line of identity at all ages. For 122 T7 ( Figure 1c) the fitted line was below the line of identity from age 50 years. However, the δ-age 123 gap KYH-T7 was positive across the entire population, as shown by the fitted line for KYH being 124 above that of T7 for the entire range of chronological age shown, the gap being slightly larger at 125 older ages. There were small differences in δ-age between men and women in KYH (women 126 were predicted 0.3 years younger on average, p=0·28) and T7 (0·3 years, p=0·15).

135
-0·22) for LDL/HDL ratio, reinforcing the notion of δ-age having increased power to cross-136 sectionally differentiate between populations presenting higher CVD risk than established CVD 137 risk factors.

142
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted September 14, 2021. ; https://doi.org/10.1101/2021.09.09.21263337 doi: medRxiv preprint Finally, we adjusted for a larger set of established CVD-related risk factors which also showed 143 significant differences between KYH and T7 (Table S1). These were diastolic blood pressure, 144 pulse rate, haemoglobin A1c (HbA1c) and education level. This further adjustment reduced the 145 gap in δ-age between studies to six years difference (95% CI: 5·7, 6·3; p<0·0001), with a slight 146 reduction in the gap when further adjusting for history of hypertension and diabetes and the final 147 model accounted for 25·8% of the gap in δ-age between studies (Table 1).

151
These results confirm our hypothesis that mean difference between predicted age and 152 chronological age (δ-age) in the study populations in Russia was substantially different to the 153 difference in the Norwegian study, consistent with the much higher CVD mortality rates in Russia 154 compared to Norway. Moreover, relative to the US clinical reference population mean δ-age was

158
As discussed elsewhere, CVD mortality rates in the Norwegian and Russian cities from which the 159 study participants were drawn were very close to their respective national averages. 12

160
Showing that mean δ-age is so much higher in the Russian compared to Norwegian study 161 population strongly suggests that δ-age may be a robust biomarker of relative vascular aging

167
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted September 14, 2021. ; https://doi.org/10.1101/2021.09.09.21263337 doi: medRxiv preprint Adjustment for a wide range of CVD risk factors attenuated the difference between KYH and T7 168 by just under 20% (from 6·95 to 5·64 years). This is consistent with recent work that suggests 169 that a core set of established CVD risk factors were able to explain only a quarter of the observed 170 differences in CVD mortality between Russia and Norway at ages 40-69 years. 13 There are a 171 number of explanations for the inability of cross-sectional risk factor profiles to explain differences 172 in CVD mortality per se that may also apply to the moderate attenuation of differences in δ-age.   The specific ECG features used by the δ-age algorithm remain unclear. Recent advances have 191 been able to highlight the segments of an ECG that a CNN algorithm uses to make predictions 14 .

192
However, the unresolved functioning of the algorithm does not prevent the objective evaluation of 193 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.    We used linear regression models to assess association of the risk factors as exposures using 234 age as the outcome. All models were adjusted for a priori confounders including chronologic age 235 and sex. True chronologic age was included in every model to remove the effects of any potential 236 correlation between δ-age and chronologic age. A test for trend for smoking was performed by 237 converting the ordered categories of number of cigarettes smoked into integer values. Basic 238 models including one risk factor at a time were further adjusted for potential confounders.

239
Density plots were generated for δ-age, SBP, BMI and Ratio LDL/HDL. In order to assess the

272
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint  (T7) and Know Your Heart (KYH) studies, chronological age was obtained via questionnaire data 336 and ECG-age was obtained by performing an, at least, ten second resting 12-lead ECG and 337 processing the ECG raw data through a CNN model trained on a USA clinical population to 338 predict age. δ-age was obtained by subtracting the chronologic age to the ECG-age. b)

339
Scatterplot of ECG-age versus chronological age for 3,487 participants form KYH (Russia) in blue . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted September 14, 2021. ; https://doi.org/10.1101/2021.09.09.21263337 doi: medRxiv preprint