Electrocardiogram abnormalities and cardiovascular risk prediction in older Chinese: the Guangzhou Biobank Cohort Study.

Background: Evidence on the associations of multiple minor ECG abnormalities (EA) with cardiovascular disease (CVD) and mortality in older populations is limited, particularly whether a weighted EA score better predicts CVD risk than a single EA severity.

Methods: We analysed 26 846 Chinese aged 50+ years from Guangzhou Biobank Cohort Study (GBCS), without CVD at baseline. Minor and major EAs were classified based on the Minnesota Code Manual. EA severity was defined as normal, one minor, two or more minor and major abnormalities. Cox regression with backward stepwise selection was conducted to develop EA score. Cox regression was used to examine the associations of EA (severity/score) with incident CVD events, all-cause mortality and CVD mortality. C-index and Net Reclassification Index (NRI) were used to assess the improvement in CVD risk prediction after adding EA (severity/score) to the GBCS model variables.

Results: During an average follow-up of 15.3 (SD=3.5) years, 6232 CVD events and 5960 deaths occurred. Compared with normal ECG, one minor (adjusted HR 1.12, 95% CI 1.05 to 1.19), two or more minor (1.20, 95% CI 1.11 to 1.29) and major abnormalities (1.46, 95% CI 1.31 to 1.63) were associated with a higher risk of incident CVD events. The EA score showed a strong dose-response relationship (0 point as reference): 1-29 points (1.12, 95% CI 1.05 to 1.19), 30-59 points (1.56, 95% CI 1.38 to 1.77), ≥60 points (3.16, 95% CI 2.56 to 3.91) (p value for trend <0.001). Similar findings were observed for all-cause and CVD mortality. Adding EA score improved the C-index for incident CVD events, but the improvement diminished over time (change in C-index: 0.011 (95% CI 0.002 to 0.022) at 3 years to 0.003 (95% CI 0.002 to 0.004) at 15 years). The NRI for 10-year risk was 0.016 (95% CI 0.007 to 0.024), indicating limited utility.

Conclusions: Major EA and multiple minor EAs were associated with higher risks of CVD events and mortality, but the value in improving CVD risk prediction is limited.