The aim of this study was to provide a comprehensive assessment of long-term associations between cardiac autonomic function and multiple health outcomes in individuals without cardiovascular disease using HRV indices measured in 15-second ECGs (usHRV). The main findings are: (1) Agreement between ultra-short (15-second) and short-term (6-minute) HRV indices was very good for RMSSD and SDSD, and moderate for PHF. (2) Low usHRV was associated with increased long-term risk of AF, MACE, stroke, and mortality, independently of resting heart rate and standard cardiovascular risk factors. In adjusted models, reduced usHRV carried a similar long-term risk for AF, MACE, and mortality as standard risk factors such as diabetes, hypertension, and smoking.
Our analysis of 1,337 NSHD participants showed that the agreement between 6-minute and 15-second HRV indices was very good for RMSSD and SDSD and good for PHF. We also demonstrated that the use of 15-second instead of 6-minute HRV did not substantially impact on regression models assessing association with prevalent CVD (data on incident CVD were not available in NSHD). A higher agreement for RMSSD and SDSD is expected because these are based on successive differences of normal heartbeats and capture fast changes in the heart rate, while PHF, which assesses respiratory-related oscillations with period ranging between 2.5 and 6.7 seconds, may be affected by reduced spectral resolution when resting heart rate is particularly low.
While the prognostic value of reduced HRV is well established in patients with cardiovascular disease, mainly post-myocardial infarction 16, only a few studies have investigated its prognostic value in population-based cohorts 4–8,17,18. Most of these studies focused on one single outcome and derived HRV from longer ECG recordings (24 hours 18, 2 hours 5,17 and 2 minutes 6,7) with only two studies using 10–15 second ECGs 8,9, but in smaller cohorts.
An association between HRV and incident AF was found in previous, smaller, studies 7,8,19. Two studies reported U-shaped associations between HRV and incident AF 7,8, which were not confirmed by our data. In Supplementary Figure S9, we demonstrate that an apparent association between high usHRV and AF can be produced when just few individuals (accounting for 0.2%-0.6% of the study population) without known cardiovascular disease but with premature atrial contractions in the 15-second ECG were included in the analysis. This suggests that premature atrial contractions, which are strongly associated with incident AF 20 and dramatically, but artifactually, increase HRV if not removed, may explain previously reported association between high HRV and AF. Our data however showed a U-shaped association between RMSSD and SDSD, and MACE and mortality even after rigorously including only individuals in normal sinus rhythm. Similar findings were reported in the Rotterdam study 9 and further investigation is required to clarify this association. Data on HRV and incident stroke is also limited, with only few studies reporting associations in population based cohorts 21,22. Our findings should encourage further investigation on autonomic dysfunction and risk of stroke in individuals without known cardiovascular disease.
The fact that significant associations were found for different health outcomes can be explained by the role that the autonomic nervous system plays in multiple aspects of cardiac and cardiovascular function. The mechanisms whereby autonomic dysfunction increases the risk of AF, MACE, stroke, and mortality are not completely understood and require further investigation.
The use of usHRV indices derived from 15-second ECGs may open new opportunities for autonomic nervous system assessment at scale, because it can be measured from standard clinical ECGs, which are taken in millions of individuals every day. This number is expected to grow dramatically thanks to popular wearable devices, including smartwatches and mobile Apps 23, which usually record the ECG for 10–30 seconds. Interestingly, in this study, usHRV showed a similar association with MACE and mortality as HRR, an established cardiac autonomic marker and risk-predictor which however requires a standardised exercise stress test. Although our findings may have a limited impact on patient-specific clinical care, 15-second HRV can improve our understanding of CVD mechanisms at the population level, including providing biological insight into the interaction and causal links between cardiac autonomic dysfunction and CVD 24.
This study has several strengths. It included both cross sectional analyses to validate usHRV indices and longitudinal analysis of a large prospective cohort (UKB, n = 51,628) for investigating interaction with outcomes. The follow-up period was long (over 11 years), and we used linear and non-linear regression models adjusted for multiple risk factors to study associations with multiple outcomes. ECG data analysis used state of the art signal processing, and visual inspection of 34,561 ECGs (61% of total) by expert reviewers in a process designed to accurately identify abnormal ECGs (estimated 0% false positive rate and 0.06% false negative rate) 20.
Several limitations need to be acknowledged. Several HRV indices 1, e.g. SDNN, cannot be measured from 15-second ECGs. Hospital episode statistics may underestimate the true incidence of outcomes and some delay may exist between an event and the date of its reporting. There is evidence of “heathy volunteer” selection bias in the UK Biobank, which may not be representative of the general population. Due to the limited number of events, myocardial infarction, heart failure and life-threatening ventricular arrhythmia were combined in the aggregate outcome MACE. However, a significant association between reduced uHRV and incident heart failure or myocardial infarction, the main components of MACE, was also found.