Baseline and 10-year change in the number of ideal cardiovascular health metrics and sudden cardiac death in the community

Abstract

Aims

Adherence to an ideal cardiovascular health (CVH) might contribute to lower the burden of sudden cardiac death (SCD) in the community. We aimed to examine the association between the number of ideal CVH metrics at baseline and of its change over 10 years with the risk of SCD.

Methods and results

The Copenhagen City Heart Study is a community-based prospective cohort study. The number of ideal CVH metrics (range 0–6; non-smoking and ideal level of body mass index, physical activity, untreated glucose, untreated systolic blood pressure, and untreated total cholesterol levels) at baseline in 1991–94 and its 10-year change thereof between 1981–83 and 1991–94 were evaluated. Definite SCD was defined as a death occurring within 1 h (eye-witnessed case) or within 24 h (non-eye–witnessed) of symptoms onset, with the presence of confirmed ventricular tachycardia and the exclusion of non-cardiac cause at autopsy. Fine and Gray sub-distribution HRs (sHRs) were calculated to account for competing risk. The study population includes 8837 participants (57% women; mean age 57 years, ±15 years) in 1991–94. After a median follow-up of 22.6 years from 1 January 1993 up to 31 December 2016, 56 definite SCD occurred. The risk of definite SCD decreased gradually with the number of ideal metrics in 1991–94 [sHR = 0.58; 95% confidence interval (CI): 0.44–0.75 per additional ideal metric] and with the change (i.e. improvement) in the number of ideal metrics between 1981–83 and 1991–94 (sHR = 0.68; 0.50–0.93 per change in the number of ideal metrics). Effect size was lower for coronary death, all-cause mortality, and coronary heart disease events.

Conclusion

Adherence to a higher number of ideal cardiovascular health was related to a substantial lower risk of definite SCD.

Graphical Abstract

Open in new tabDownload slide

Introduction

In US and European adults, 185 000–450 000 and 155 000–345 000 cases of sudden cardiac death (SCD) occur each year, with over a half in asymptomatic individuals, making prevention of SCD in the community a public health priority.^1,2 Despite the identification of specific risk factors,^3,4 genetic variants,⁵ symptoms prior to SCD,^6,7 or the development of risk prediction algorithms,^8,9 we currently fail to prevent SCD in the community.¹⁰ Alternative preventive strategies are therefore needed. In addition to controlling risk factors once they are in place (primary prevention), preventing risk factor development may be of importance (primordial prevention). One possibility to avoid risk factors appearance could be to adhere and maintain a cardiovascular health (CVH) at the ideal (i.e. optimal) level throughout life. This concept has been promoted by the American Heart Association (AHA) which defines the CVH score also known as Life’s Simple 7, a 7-item score comprised of four behavioural metrics (non-smoking and ideal level of body mass index, physical activity, and diet) and three biological metrics (ideal levels of untreated blood pressure, fasting plasma glucose, and total cholesterol).¹⁰ Many observational studies demonstrated that adherence to a higher CVH score and change in CVH score over time were both related to a lower risk of cardiovascular disease (CVD) and mortality.^11–13 Because smoking, obesity, physical inactivity, low consumption of fish, nuts, or polyunsaturated fat, high blood pressure, and diabetes have been each related to a higher risk of SCD, a strategy aimed at preventing the joint exposure to these risk factors might be relevant to prevent SCD in the community.^3,14–18 So far however, only two studies have specifically assessed the association between CVH score and SCD risk in the community, one examining CVH at baseline exclusively in men¹⁹ and another addressing change in CVH over a short period (6 years).²⁰ The aim of this study was therefore to extend current evidence on the possible association between CVH score and SCD in the community, by quantifying the relationship between baseline CVH score and change in CVH score over 10 years with subsequent risk of definite SCD in adult men and women from the community. To contextualize the findings, associations with all-cause mortality, total SCD (definite, probable or possible SCD), non-sudden coronary death, and coronary heart disease (CHD) events were further examined. These questions were addressed in the Copenhagen City Heart Study (CCHS) within the framework of the European Sudden Cardiac Arrest network: towards Prevention, Education, and New Effective Treatments (ESCAPE-NET) network.^21,22

Methods

Study design

The CCHS study was initiated in 1976 and recruited a random age- and sex-stratified sample of individuals in the general population aged 20–93 years living in a district (Østerbro) of the Danish capital, Copenhagen. Physical examination took place in 14 223 participants (73.5% participation rate) in 1976–78 and subsequently in 1981–83, 1991–94, and 2001–03.²² Follow-up for SCD started on 1 January 1993 when death certificates became digitized and ended up in 31 December 2016, meaning that SCD that occurred prior 1 January 1993 could not be accounted for. Therefore, to minimize survival bias, baseline CVH score was computed using data collected in 1991–94 whereas change in CVH score over 10 years was computed using data collected between 1981–83 and 1991–94 (main analysis) (see Supplementary material online, Figure S1). The Regional Ethical Committee for Medical Research in Copenhagen approved the original study. All the participants gave informed consent at enrolment.

Definition of the cardiovascular health score

Six out of the seven CVH metrics (information on diet was missing) were available at each examination round. Following the AHA criteria, the ideal level of each metric was defined as follows: body mass index (BMI) < 25 kg/m², never smoking or smoking cessation for more than 1 year, physical activity more than 180 min per week (intensity of the physical activity was not available), total cholesterol in target range (<200 mg/dL) without treatment, blood pressure in target range (<140/90 mmHg) without treatment, and absence of physician-diagnosed diabetes (fasting glycaemia level was unavailable).¹¹ Each metric was assigned a score of 1 if at ideal level and 0 otherwise, so that the CVH score corresponding to the number of ideal metrics ranged from 0 to 6. Main exposures included the CVH score at baseline in 1991–94 (0–6) and the absolute change in CVH score 10 years prior 1991–94, i.e. between 1981–83 and 1991–94 (range −4 to +3).

Data sources

All Danish citizens are issued a unique personal identification number which is linked across all Danish national registries and used in the CCHS. This study used data from (i) the Danish Cause of Death Register, which contains information regarding the cause and circumstances of deaths since 1970 including the original death certificates (detailed in the Supplementary material), and (ii) the Danish National Patient Register, which contains information on all inpatient activities since 1977 and outpatient visits since 1994. In particular, the immediate, contributory, and underlying causes of death are registered with a possibility to specify the timing of the causes, which makes Danish death certificates highly suitable for the identification of sudden and unexpected death.²³ All available hospital discharge summaries and autopsy reports from forensic departments and pathology departments were included and accessed digitally.

Definition of sudden cardiac death

The reviewing process and definition of SCD (i.e. definite, probable, possible SCD) has been reported previously.²⁴ All deaths were reviewed case by case using all available information present in the above-mentioned data sources. Two physicians reviewed all information independently to identify all potential SCDs. In case of disagreement, a third reviewer was involved to re-evaluate the case and reach a consensus. Sudden cardiac death was defined as a sudden, unexpected, and natural death, in witnessed cases as an acute change in cardiovascular status leading to death <1 h and for unwitnessed cases as a person seen alive and functioning normally <24 h before being found dead. Definite SCD, which is the main outcome of the present study, was defined as a person with an autopsy with cardiac or unknown cause of death, or a death with confirmed ventricular arrhythmia preceding death, with an established time frame from change in cardiovascular status to death. Probable SCD was defined as a presumed cardiac origin in non-autopsied cases after review of all available information with an established time frame from change in cardiovascular status to death. Possible SCD was defined as a presumed cardiac origin in non-autopsied cases after review of all available information without a fully established time frame from change in cardiovascular status to death. Only out-of-hospital SCD was included in this study as it represents the vast majority of SCD. Total out-of-hospital SCD included definite, probable, or possible out-of-hospital SCD.

Definition of non-sudden cardiac death cardiovascular events

The CHD events were defined as International Classification of Diseases (ICD)-10 codes I20 to I25 or Z951 or Z955. Coronary death was defined as ICD-10 codes I20 to I25.²³

Covariables

Covariates included marital status (married/cohabiting with a partner or not), education level (more or less than 12 years education), and depression (physician diagnosis and participants reporting taking nerve medication or sedatives daily or almost daily).

Statistical analysis

Main analyses

Main analysis examined the association of baseline CVH score in 1991–94 and of change in CVH score between 1981–83 and 1991–94 with subsequent definite SCD. Sub-distribution hazard ratios (sHRs) were calculated per number of ideal CVH metrics in 1991–94 and per change in the number of ideal CVH metrics between 1981–83 and 1991–94 using competing risk analysis (Fine and Gray). In the main analysis, non-SCD death was used as competing event.²⁵ Models used age as the timescale were stratified on birth date (5-year interval) to account for a birth cohort effect and were adjusted for sex, education, marital status, and diagnosed depression, as these covariates have been previously related to both SCD^26–29 and the CVH score.^30,31 Analysis was further adjusted for the number of ideal CVH metrics in 1981–83 when examining change in CVH between 1981–83 and 1991–94. Methods to assess the proportional hazard assumption and log linearity assumption are detailed in the Supplementary material. Population-attributable fraction (PAF) was calculated to quantify the proportion of definite and total SCD that could potentially be avoided in this cohort according to the number of ideal metrics achieved. The PAF estimations account (i) for the prevalence of the number of ideal CVH metrics achieved in 1991–94 and the related multivariate cause-specific HRs, using those with 0–1 ideal metric as the reference category, (ii) for the prevalence and the related multivariate cause-specific HRs for the covariates of the model, and (iii) for the prevalence and HRs of unmeasured risk factors (see Supplementary material).³²

Secondary analyses

To contextualize the main findings, association of the baseline CVH score in 1991–94 and of its change between 1981–83 and 1991–94 with total SCD (i.e. definite, probable, and possible), incident CHD events, non-sudden coronary death, and all-cause mortality was quantified, following the same methodology as for the main analysis and using respectively all-cause death and non-cardiovascular death as competing events.

Sensitivity analyses

This section is detailed in the Supplementary material and includes (i) cause-specific HR estimations, (ii) inverse probability weighting analyses to account for sample attrition,³³ (iii) an analysis by individual CVH metric, and (iv) an analysis of change in CVH score over 15 years between 1976–78 and 1991–94. The analysis of change in CVH score over 10 years between 1991–94 and 2001–2003 examined total SCD as an outcome since only seven definite SCD occurred after 2001–03. To further elaborate on the clinical relevance of the study findings, the possible association between CVH score and resting heart rate, a strong risk marker for SCD,^3,4 was evaluated. More specifically, we run linear regression analyses to relate CVH score 1991–94 and of its change between 1981–83 and 1991–94 with resting heart rate in 1991–94 on the one hand and with the average of resting heart rate measured in 1991–94 and in 2001–03 on the other hand.

Statistical analysis was conducted using SAS® 9.4 (SAS Institute Inc., Cary, NC, USA).

Results

Baseline characteristics in 1991–94

Of the 10 135 participants examined in 1991–94, after excluding those who died between 1 January 1991 and 31 December 1992 before the start of death certificates digitization (n = 35), those with prior CVD (n = 679), missing information on the CVH metrics (n = 506), and covariates (n = 78), the study population included 8837 participants [57% women; mean age of 57 years (SD 12)]. As shown in Table 1, blood pressure was the least prevalent ideal metric (15%) whereas absence of physician-diagnosed diabetes was the most prevalent one (97%). The proportion of ideal smoking status, ideal BMI, and ideal blood pressure was higher in women, whereas the opposite was seen for ideal physical activity and ideal total cholesterol; no sex disparities were seen for the distribution of the diabetes status.

Cardiovascular health score in 1991–94 and subsequent risk of sudden cardiac death

After a median follow-up of 22.6 years from 1 January 1993 up to 31 December 2016, a total of 4610 died, including 56 definite SCD, 201 probable SCD, and 371 possible SCD (n = 628 total SCD, representing 13.6% of all deaths). Mean age at definite SCD onset was 73.69 (11.49) compared to 80.13 (10.00) for total SCD, 81.29 (9.34) for non-sudden coronary death, and 70.21 (11.09) for CHD events, respectively. Higher CVH score was associated with a significantly lower risk of definite SCD (sHR = 0.58; 95% CI: 0.44–0.75 per additional ideal metric) after adjusting for sex, education, marital status, and depression diagnosis in 1991–94. Effect size was lower albeit statistically significant for total SCD, all-cause mortality, non-sudden coronary death, and CHD events, with sHR ranging from 0.76 (95% CI: 0.70–0.83) for total SCD to 0.82 (95% CI: 0.72–0.93) for non-sudden coronary death (Figure 1 and Graphical Abstract). The association between CVH score and definite SCD was stronger in women compared to men (sHR = 0.37; 0.24–0.57 vs. sHR = 0.74; 0.55–1.00; P for sex interaction = 0.008), whereas no sex interaction was observed for the other investigated outcomes. Association between CVH score and definite SCD did not differ by age (P for interaction = 0.24).

Population-attributable fraction

As shown in Table 2, between 19.2% (95% CI: 9.6; 28.4) and 62.2% (32.4; 80.7) of total SCD events could be potentially avoided if the population would achieve respectively 2 and 5–6 ideal metrics compared to having achieved 0 or 1 ideal metric. The PAF estimates for definite SCD were of higher magnitude but did not systematically reach statistical significance due to the small number of cases per number of ideal metrics.

Changes in the cardiovascular health score between 1981–83 and 1991–94 and subsequent risk of sudden cardiac death

The change analysis was conducted in 6156 participants (mean age 52.5 years, 59% women) who attended examination round in 1981–83 and 1991–94, were free of CVD in 1991–1994, and had the six CVH metrics and no missing covariates at both time points. The characteristics of excluded and included participants are reported in Supplementary material online, Table S1. Overall, 63% did not change their CVH score including one-third remaining with two ideal metrics or less; 25% decreased the number of ideal metrics mainly from three or four ideal metrics to two ideal metrics or less; and 11% increased the number of ideal metrics, mainly from two ideal metrics or less to three to four ideal metrics, respectively (Figure 2). No sex disparities were noted.

After a median follow-up of 18.3 years, 3979 participants died including 51 definite SCD, 178 probable SCD, and 328 possible SCD, respectively (n = 557 total SCD). As shown in Table 3, improvement in CVH score was related to a significantly lower risk of definite SCD (sHR = 0.68; 0.50–0.93 per change in the number of ideal metrics between 1981–83 and 1991–94) after adjusting for sex, education, marital status, depression diagnosis in 1991–94, and the CVH score in 1981–83. These associations did not differ by sex or by age. In this model, the CVH score in 1981–83 remained associated with definite SCD (sHR = 0.49; 0.35–0.69 per number of ideal metrics in 1981–83). For comparison and as for the baseline analysis, effect size was of lower magnitude for total SCD, all-cause mortality, non-sudden coronary death, and CHD events, with sHR ranging from 0.82 (95% CI: 0.74; 0.90) for total SCD to 0.90 (95% CI: 0.82; 0.99) for CHD events, respectively (Table 3 and Graphical Abstract).

Sensitivity analyses

Consistent findings were observed when using cause-specific analysis (although with stronger effect size) (see Supplementary material online, Table S2) and inverse probability weighting analysis (see Supplementary material online, Table S3), when examining change in CVH score over 15 years between 1976–78 and 1991–94 (see Supplementary material online, Table S4), as well as change in CVH score between 1991–94 and 2001–03 (see Supplementary material online, Tables S5 and S6 and Figures S1, S2 and S3, for total SCD). The individual metric analysis shows that ideal smoking status and absence of physician-diagnosed diabetes at baseline in 1991–94 were each related to a lower risk of definite SCD (see Supplementary material online, Tables S7 and S8) and that remaining free of physician-diagnosed diabetes over 10 years was related to lower risk of definite SCD (see Supplementary material online, Table S9). Furthermore, higher CVH score in 1991–94 was related to lower resting heart rate at the same time period (−1.83 b.p.m. difference per one additional ideal metric). Similarly, improvement in the number of ideal metrics between 1981–83 and 1991–94 was related to lower resting heart rate 1991–94 (−1.22 b.p.m. per change in the number of ideal metrics) and to lower averaged resting heart rate over 1991–94 and 2001–2003 (−1.15 b.p.m. per change in the number of ideal metrics), respectively (see Supplementary material online, Table S10).

Discussion

In this large Danish prospective population-based study, higher CVH score and increase in CVH score over 10 years were each related to a significantly decreased risk of definite SCD over 22.6 and 18.3 years of follow-up, respectively. These associations were of stronger magnitude compared to total SCD (definite, probable, or possible SCD), non-sudden coronary death, CHD events, and all-cause mortality.

Prior to this study, adherence to a healthy lifestyle (not smoking, BMI < 25 kg/m², exercise ≥ 30 min/day, and top 40% of the Alternate Mediterranean Diet Score) was strongly and inversely related to SCD in the Nurses’ Health Study, but the analysis was exclusively conducted in women and did not consider biological risk factors.³⁴ Two prior studies have specifically examined the association between CVH score and SCD risk in the community. In 2755 men aged 42–60 years from the Kuopio Ischemic Heart Disease cohort study in Finland, those with five to seven ideal metrics in 1984 had a 85% lower risk of SCD compared to those with two or less ideal metrics over 25.8 years of follow-up.¹⁹ Change in CVH score was recently examined in the ARIC study in the USA. In 12 207 men and women aged 45–64 years, compared to participants with persistently poor CVH over 6 years, those who changed from poor to intermediate/ideal, and those who changed from ideal to intermediate/poor had respectively a 27% and a 77% risk reduction of SCD over 23 years of follow-up.²⁰ The present study extends the results of the two above specific studies on the following points. The wider age range from 20 to 93 years compared to 42 to 65 years permitted to examine more precisely the influence of age. That the association between CVH score and SCD risk did not differ according to age is an important message suggesting that adherence to a higher CVH score has the potential to benefit to all segments of the (adult) population including the oldest who are at higher risk of SCD, in accordance with what has been previously observed for CHD or stroke risk.³⁵ Furthermore, the Finnish cohort was exclusively conducted in men while we additionally included women, permitting to identify a significant sex interaction. The stronger association between the baseline number of ideal metrics and SCD risk in women compared to men in the present study is consistent with the observation that women adhere more often than men to higher level of CVH.³⁶ Change in CVH score was explored over 6 years in ARIC as compared to 10 years (main analysis) and 15 years (secondary analysis) in the present study; more recent patterns of change in CVH score were also considered here (between 1991–94 and 2001–2003) compared to between 1987–89 and 1993–95 in ARIC. Finally, the inverse association between baseline and change in CVH score with resting heart rate, a strong risk marker for SCD, permitted to strengthen the clinical relevance of the study findings.^3,4

The current findings carry important public health implications. Based on our PAF analysis, achieving 2 ideal metrics (realistic approach) or 5–6 ideal metrics (idealistic approach) compared to achieving none or 1 could potentially prevent 19.2% and 62% of total SCD events, respectively. By extrapolation, between 48 000 and 155 000 total SCD cases could be theoretically avoided each year in Europe.² That the reduced risk of SCD associated with improvement in CVH was independent from the baseline CVH score is another important message suggesting that it is never too late to start changing its health behaviours no matter how low is your CVH score. On the other hand, the stronger effect size associated with baseline compared to change in CVH score confirms the importance of the initial level of CVH.¹³ Consistent with prior studies, less than 10% of the study population had at least five ideal metrics, and only 11% improved their CVH score between 1981–83 and 1991–94. More recent trends were encouraging in this study with CVH improvement almost doubling between 1991–94 and 2001–2004 as compared to a decade earlier (11% vs. 18%). An appropriate but challenging strategy to keep improving these figures could be to start promoting and monitoring primordial prevention of SCD early in life.^37–39 This approach requires a concerted effort from the individuals, their families, the school environment, caregivers, and health policy, but also strong political initiatives to influence population health and the environment in which people reside. Finally, the stronger association of the CVH score with definite SCD as compared to non-sudden cardiovascular mortality and CHD events should be viewed as a continuum of the potential benefit of primordial prevention across the spectrum of CVD. The present findings suggest that primordial prevention has the potential to be a unified approach for the prevention of CVD in general and SCD in particular.

Despite the strengths of the study including the study sample size, the longitudinal design, the length of follow-up, and the precise definition of the SCD phenotype, we acknowledge several limitations. First, the observational nature of the study precludes drawing any causal conclusions between the CVH score and SCD risk. In particular, PAF estimates should be interpreted with caution since its calculation is assuming a causal relationship. Similarly, residual confounding due to unmeasured risk factors such as anxiety,^{40, 41} family history of SCD,^3,42 alcohol consumption,⁴³ or left ventricular dysfunction⁴⁴ cannot be excluded. Also, adherence to a higher CVH score and change in CVH score over time may capture additional unmeasured factors and long-term habits that may partly contribute to the lower risk of SCD associated with higher CVH score. Second, we decided to focus on definite SCD to maximize the specificity of SCD phenotype but this was at the expense of a loss in statistical power. Third, we acknowledge misclassification bias in the definition of the CVH score. The diet metric was not available, whereas fish consumption and polyunsaturated fat are inversely related to SCD.^16,17 The intensity of physical activity was missing and the glucose metric was only based on physician-diagnosed diabetes. These classification biases are likely to underestimate the true association between the CVH score and the studied outcomes. Fourth, by design, the 35 participants who died between 1991 and 1993 (i.e. before death certificates digitization) could not be included in the analysis, but considering that SCD accounts for 20% of all-cause mortality, only seven SCD cases could have been missed in the present analysis. Furthermore, change in CVH score analysis could be exposed to selection bias owing to the number of excluded participants and to the difference in the baseline characteristics between included and excluded participants. Of note however, the results remained unchanged when the inverse weighted analysis to account for attrition was used. Fifth, the participants were of European descent and living within a Northern European urban context including a universal health coverage; thus, the results may not apply to other ethnic groups or contexts. Of note however, the association between CVH score and SCD in ARIC did not differ by race.²⁰ Finally, the Life’s Essential 8, which additionally considers sleep duration, was recently proposed,⁴⁵ but this score was not available at the time of the current analysis.

To conclude, the observation that adherence to a higher CVH score and improvement in CVH score over 10 years are related to a substantially lower risk of definitive SCD represents initial but important steps before envisaging primordial prevention as one possible relevant strategy to reduce the burden of SCD in the community. This strategy should be viewed as being complementary to ongoing examined approaches and interventions that are targeting the general population and the populations at high risk of SCD.⁴⁶

Supplementary material

Supplementary material is available at Europace online.

Authors’ contribution

Conceptualization, methodology, supervision, and writing original draft: J.P.E. Data curation, methodology, and statistical analysis: M.C.P. Writing—review and editing: all other authors. Fundings: H.T., J.T.-H., E.P., and R.J.

Funding

This project has received funding from the European Union’s Horizon 2020 research and innovation programme ESCAPE-NET under grant agreement no. 733381, and the PARQ COST-ACTION funding under grant number CA19137.

Data availability

The data used for this study cannot be shared freely. The data contain potentially identifying and sensitive patient information. The data can be accessed by contacting ESCAPE-NET’s study coordinator ([email protected]) and the leader of the CCHS ([email protected]), but cannot leave the high security data environment of the ESCAPE-NET consortium, due to their sensitive nature.

References

Author notes