Change in resting heart rate and risk for all-cause mortality

While resting heart rate (RHR) is known to have strong ties to genetics,1 there is also a large amount of variability in RHR that can be attributed to different lifestyle factors like physical activity, diet, stress, and quality of sleep.2 Consequently, RHR is an important and easily attainable proxy of cardiovascular and general health. The Framingham study in 1987, and similar cross-sectional cohort studies after that, have shown that an elevated RHR at a given time point is associated with a greater risk of all-cause and cardiovascular mortality.3 Collectively, these studies have suggested that with every 10 b.p.m. increment in RHR, risk for all-cause mortality (ACM) increases by ∼9% (95% CI 7–12%). However, while the cross-sectional cohort is ideal for identifying the relative risk of a (sub)population, such studies offer limited insight into the relative risk of an individual over time in relation to changes in RHR. Repeated measures of RHR may therefore be more appropriate in understanding the association between RHR and mortality. In that light, we conducted a systematic review (Prospero registration: CRD42020166869) to determine the risk for ACM associated with a change in RHR over time in the general population. Searches were conducted in PubMed, Web of Science, and Scopus for all relevant documents pre-dating 10 January 2021 focussing on ‘resting heart rate’, ‘mortality’, ‘risk’, and ‘change’. Longitudinal cohorts in adults, with no specific disease at the time of enrolment were considered, in which the exposure of interest (RHR) had to be measured at least twice, at a minimum time interval of 1 year. All-cause mortality had to be considered at a minimum time interval of 1 year post the most distant (in time) RHR measurement. Study selection and data extraction were conducted by two reviewers independently (B.R. and M.H.). Random-effect meta-analyses were conducted (Review Manager 5.3) based on a generic-inverse variance model, stratified by the direction of the potential change in RHR (stable vs. increase; stable vs. decrease). Eleven studies (Table 1; 108 625 participants) from seven different countries were identified, from 1287 unique articles, of which 9 could be included in the meta-analyses.4–13 The mean time window between RHR measurements was 6.3 (±4.9) years while the average follow-up period for determination of ACM was 11.3 (±4.7) years. The weighted average cut-off for an ‘increase’ in RHR was 2.6 b.p.m. (±2.8 b.p.m.); albeit most studies used >5 b.p.m. to separate stable from increasing. The cumulative hazard ratio (see Figure 1) for ACM associated with an increased RHR was 1.21 (95% CI 1.09–1.35, P < 0.01). When comparing a stable RHR against a decreasing RHR, seven of the nine studies provided sufficient data for analysis, reporting on 57 846 participants. The weighted average cut-off for a ‘decrease’ in RHR was −4.48 b.p.m. (±1.50 b.p.m.). The cumulative hazard ratio (see Figure 1) for ACM associated with a decrease in RHR was 0.96 (95% CI 0.87–1.05, P = 0.31). Forest plot of the risk ratio associated with an increasing resting heart rate relative to a stable heart rate (top) as well as the risk ratio associated with a decreasing resting heart rate relative to a stable heart rate (bottom). Details of the 11 studies included ACM, all-cause mortality; AF, atrial fibrillation; BMI, body mass index; BP, blood pressure; ECG, electrocardiogram; eGFR, estimated glomerular filtration rate; NT-pro-BNP, N-terminal pro-b-type natriuretic peptide; RHR, resting heart rate; Unk, Unknown. Time between 1st and last RHR measure. Time between last RHR measures and assessment of ACM. Details of the 11 studies included ACM, all-cause mortality; AF, atrial fibrillation; BMI, body mass index; BP, blood pressure; ECG, electrocardiogram; eGFR, estimated glomerular filtration rate; NT-pro-BNP, N-terminal pro-b-type natriuretic peptide; RHR, resting heart rate; Unk, Unknown. Time between 1st and last RHR measure. Time between last RHR measures and assessment of ACM. Collectively, the results of this review may indicate that an increase in RHR is associated with a higher risk (9–35%) for ACM, while a decrease in RHR did not lead to a statistically significant ‘protective’ effect or risk reduction. Moreover, these findings highlight the potential value of monitoring RHR over time at an individual level (i.e. within-person), rather than a population level (i.e. between-person). While previous cross-sectional cohort studies have suggested a higher risk for ACM for individuals with a 10 b.p.m. increase in RHR,4 the results in this review suggest that a within-person increase (based on repeated measures) as low as 2.6 b.p.m. (±2.8 b.p.m.) may already be associated with a significant increase in ACM. Yet, the underlying driving factors of change in RHR are likely plentiful, complex, and interrelated; including both intrinsic (e.g. behaviour) and extrinsic factors (e.g. air quality) that affect the biological response.15 Hence, while the use of repeated measures eliminates some between-person heterogeneity, a change in RHR can only be interpreted in conjunction with other measures. However, while cognizant of these limitations in identifying causal mechanisms, one could argue for the use of RHR as an accessible biomarker for general health, and in case an increase in RHR beyond a certain threshold is detected, this could warrant a more aggressive approach to lifestyle (e.g. smoking, physical inactivity), clinical risk factor screening (e.g. stress electrocardiogram) and/or intervention. Based on this review, the weighted average increase in RHR was 2.6 (±2.8) b.p.m.; however, most studies used >5 b.p.m. as a conservative threshold to differentiate stable from increased. One of the key limitations in terms of RHR assessment may be a lack of standardization in terms of its assessment and statistical modelling. The psychometric properties (e.g. criterion validity, test–retest reliability) of RHR in research, and more importantly, routine clinical practice will largely depend on the level of standardization between assessments, as well as the opportunity to account for potential confounders. Future research is needed that elucidates such properties specifically,1 including reliability and minimal detectable change in a controlled and clinical environment, to improve the value of (change in) RHR in clinical decision-making. Furthermore, widely accepted recommendations for the assessment of RHR, in terms of procedures and supporting variables, for repeated assessment specifically, could assist. A second important area for future research is related to wearable technology, and the implementation thereof in routine health care.16 In particular in settings where wearable device technology is widely available, there is sufficient argument for tracking of RHR over time using a more ecological and momentary approach, and potentially, allow for early detection of changes in health state. Multiple assessments over shorter time windows may allow increased sensitivity, as well as recall of factors that coincided with a change in RHR once detected. Independent of access to wearable technology,17 addressing the limitations in RHR assessment over time is paramount for RHR to be fully adopted as an easy accessible clinical biomarker for general health. Conflict of interest: none declared.