|Home | About | Journals | Submit | Contact Us | Français|
Sudden cardiac death (SCD) accounts for over half of all cardiac deaths, and the majority occur as the first manifestation of heart disease, especially among women. Primary preventive strategies are needed to reduce SCD incidence.
To estimate the degree to which adherence to a healthy lifestyle may lower risk of SCD among women.
We conducted a prospective cohort study among 81,722 US women from the Nurses' Health Study, from June 1984 to June 2010. Lifestyle factors were assessed via questionnaires every 2-4 years. A low-risk lifestyle was defined as not smoking, BMI <25 kg/m2, exercise ≥30 minutes/day, and top 40% of the Alternate Mediterranean Diet Score, which emphasizes high intake of vegetables, fruits, nuts, legumes, whole grains, fish and moderate intake of alcohol.
SCD defined as death occurring within 1 hour of symptom onset without evidence of circulatory collapse.
We documented 321 cases of SCD over 26 years. All 4 low-risk lifestyle factors were significantly and independently associated with lower risk of SCD. The absolute risks of SCD were 22, 17, 18, 13 and 16 cases/100,000 person-years among women with 0, 1, 2, 3 and 4 low-risk factors, respectively. Compared to women with 0 low-risk factors, the multivariate relative risk of SCD was 0.54 (95%CI: 0.34, 0.86), 0.41 (95%CI: 0.25, 0.65), 0.33 (95%CI: 0.20, 0.54) and 0.08 (95%CI: 0.03, 0.23) for women with 1, 2, 3, and 4 low-risk factors, respectively. The proportion of SCD attributable to smoking, inactivity, overweight, and poor diet was 81% (95%CI: 52%, 93%). Among women without clinically diagnosed CHD, the PAR% was 79% (95%CI: 40%, 93%).
Adherence to a low-risk lifestyle is associated with a low risk of SCD and may be an effective strategy for the prevention of SCD in the general population.
Sudden cardiac death (SCD) accounts for over half of all cardiac deaths, with an incidence of ~250,000 – 310,000 cases in the U.S. annually1, 2. Although coronary heart disease (CHD) underlies most SCD events, SCD is the first manifestation of CHD for the majority of victims3, particularly among women4. Efforts aimed at primary prevention of SCD focus primarily on placement of implantable cardioverter–defibrillators (ICDs) in patients with severe left ventricular dysfunction. However, only 25-30% of SCDs occur among this high-risk subgroup, and therefore these efforts do not address the majority of SCDs.3 Prevention strategies applicable to lower risk populations are needed to reduce SCD.
Modifiable lifestyle factors, such as smoking,5-7 obesity,4, 5, 7 and physical inactivity6, 8, 9 are independent risk factors for SCD, and several dietary factors, including low intake of N-3 fat10-12 alcohol abstinence and heavy alcohol intake are associated with higher risk of SCD.13, 14 A low-risk lifestyle, including combinations of not smoking, prudent diet, regular exercise and healthy weight has been associated with lower risk of CHD,15 stroke,16 diabetes,17 cancer,18 hypertension19, chronic disease,20 and total and CVD mortality21, 22 and may provide an alternative means to prevent SCD. However, no studies have examined the combination of multiple lifestyle factors and risk of SCD. Therefore, we estimated the burden of SCD that potentially may be attributed to adverse lifestyle factors in the Nurses' Health Study (NHS).
The NHS began in 1976 when 121,700 female nurses aged 30 to 55 provided information on lifestyle and medical history.23 Information on newly diagnosed disease, lifestyle and other risk factors is collected repeatedly throughout follow-up. The institutional review board at Brigham and Women's Hospital approved the research protocol. Participants first completed an expanded food-frequency questionnaire (FFQ) in 1984, which served as the baseline year for this analysis; we excluded women who left ≥70 food items blank on the FFQ or reported energy intake <600 or >3500 kcal/day, leaving 81,722 women for analysis.
We obtained information biennially on smoking status, weight, menopausal status, use of medications, and physician diagnosis of disease. Information on height was collected on the baseline questionnaire and parental history of myocardial infarction (MI) was obtained in 1976 and 1984. Physical activity was assessed every 2-4 years using a validated questionnaire.24 We calculated the hours per week spent in leisure-time activities of moderate (3-6 METs/hr) or vigorous (6+ METS/hr) intensity. Specific activities included brisk walking (≥3 mph), jogging, running, bicycling, swimming, tennis, squash, racquetball, rowing, calisthenics and yoga.
A validated FFQ25 was completed in 1984, 1986 and every 4 years thereafter. For each food item, participants were asked how often, on average, a specified portion was consumed over the past year. Nutrient intake was calculated by multiplying the nutrient content of each food item26 by frequency of intake and summed across all food items. All nutrients were adjusted for total energy intake by regressing nutrient intake on total energy.27
We considered 4 lifestyle factors, smoking, exercise, diet and weight. We selected these factors based on their associations with SCD, and current recommendations for overall CVD prevention28. For each lifestyle factor, a woman received 1 point if she met the criteria for low-risk and 0 if otherwise.
For smoking, we defined low-risk as not currently smoking given strong relationships between active smoking and SCD in prior studies.5-7 Regular participation in moderate to vigorous intensity exercise has also been associated with lower risk of SCD, 8, 9 and we defined low-risk for exercise as ≥30 minutes/day of moderate or vigorous intensity activity, based on current exercise recommendations.29 Obesity is an independent risk factor for SCD,5 and a 3.3 kg/m2 increase in BMI was associated 20% higher risk of SCD.7 We defined optimal weight as BMI <25 kg/m2, which the standard cutoff point for overweight as defined by the WHO.30
For diet, we defined low-risk as an alternate Mediterranean (aMed) diet score31 in the top 40% of the cohort distribution. The aMed score quantifies adherence to a Mediterranean-style diet, which emphasizes high intake of foods and nutrients that have been associated with lower risk of SCD, including N-3 fats (GISSI), nuts32, fish10, 33 and moderate alcohol consumption13, 14. Specifically, the aMed score includes 9 components: high intake of vegetables, fruits, nuts, whole grains, legumes, fish and ratio of monounsaturated to saturated fat, moderate intake of alcohol, and low intake of red/processed meat. Women received 1 point for intake above and 0 points for intake below the median, except for red/ processed meat, where the score was reversed. Women received 1 point for moderate alcohol intake (0.5 to 1 drink/day) and 0 otherwise. Scores for the aMed score ranged from 0-9, with higher scores representing greater resemblance to the Mediterranean-style diet.
We summed the total number of low-risk factors to create a binary lifestyle score (ranging from 0 to 4). Results were similar when we assigned weights to each low-risk factor, based on the beta-coefficients from multivariable regression models with SCD as the outcome. The binary variables do not account for the gradient in risk of SCD with more extreme levels of these lifestyle factors. Thus, we calculated an expanded lifestyle score based upon the associations between each lifestyle factor and SCD in the cohort. We assigned scores of 1 (least healthy) to 5 (most healthy) to categories of the lifestyle factors and summed the points across all four factors, with possible scores ranging from 4 to 20. The assigned points for each category are presented in Table 2. For this analysis, healthiest behavior was defined as never smoking, BMI between 21-25 kg/m2, exercise ≥6 hours/week and the highest quintile of the aMed score.
We attempted to confirm each cause of death through review of medical records and autopsy reports4. If circumstances surrounding the death were not documented adequately, we ascertained additional details through interviews with next of kin. We did not rely on the death certificate to determine timing of death. Cardiac deaths were considered sudden if the death or cardiac arrest occurred within 1 hour of symptom onset. To increase the specificity for an “arrhythmic death”, we excluded women with evidence of circulatory or neurologic impairment before death.34 We included unwitnessed deaths that could have occurred within 1 hour of symptom onset, and that had autopsy findings consistent with SCD in the analysis (i.e. acute coronary thrombosis or severe coronary artery disease without myocardial necrosis or other pathologic findings to explain death; n=35). The autopsy rate in this cohort (7%) was comparable to national autopsy rates (6%).35
Women contributed person-time from the date of return of the 1984 questionnaire until incident SCD, death, or June 1, 2010, whichever came first. We hypothesized that exercise, BMI and smoking would have transient effects on SCD risk, and these variables were updated at each questionnaire cycle to reflect the most recent information. Although physical activity was assessed repeatedly during follow-up, it was not assessed on the 1984 questionnaire, thus the mean of activity in 1980 and 1982 represented physical activity at baseline in 1984. We used the cumulative average of the diet score assessed every 2-4 years to represent long-term diet and reduce measurement error.27 Because changes in diet after an intermediate endpoint such as hypercholesterolemia, diabetes, hypertension, transient ischemic attack or CHD (nonfatal MI, angina or coronary revascularization) may confound the association between long-term diet and SCD,36 we stopped updating dietary information when a woman developed an intermediate end point during follow-up. If data were missing at a given time point (13% alcohol, 12% exercise, 8% BMI, 1% smoking, 0% cumulative diet from ≥1 questionnaire), the last observation was carried forward for 1 cycle.
To estimate the association between low-risk lifestyle factors and risk of SCD, we used multivariable Cox proportional hazards models to estimate hazard ratios as estimates of the relative risk and 95% confidence intervals (CI), stratifying by age (months), and adjusting for family history of MI before 60 (yes/no), postmenopausal status (yes/no), current hormone therapy (yes/no) and history of disease (hypertension, hypercholesterolemia, diabetes, cancer, CHD and stroke) at baseline. Further adjustment for medication use (aspirin, digoxin, blood pressure lowering, cholesterol-lowering and anti-arrhythmic medications) did not appreciably alter the results. All covariates were updated each questionnaire cycle and included as time-varying covariates in multivariate models, except for family history of MI. The proportional hazards assumption was not violated.
We conducted a Wald test for linear trend by assigning the median value to each category and modeling this variable as a continuous variable. For smoking, we used an ordinal variable. For BMI and the lifestyle risk scores, we examined potential non-linear relationships with risk of SCD using restricted cubic spline transformations, to model these relationships without prior specification of the risk function.37 We conducted tests for non-linearity using the likelihood ratio test, comparing the model with only the linear term to the model with the linear and cubic spline terms.
We calculated the population attributable risk percent (PAR%) and 95% CI to estimate the proportion of SCD in this cohort that hypothetically would not have occurred if all women were in the low-risk group, assuming a causal relation.38 For these analyses, we compared women in the low-risk category (for each factor individually and in combination) with the rest of the women in the population.39 To calculate the PAR%, we estimated the relative risk from multivariate pooled logistic regression models to allow for the direct inclusion of age in the model. In this approach, each 2-year interval is treated as an independent follow-up study and observations over all intervals are pooled into a single sample. When the disease is rare, estimates from pooled logistic regression models approximate estimates from Cox proportional hazards models.40 When calculating the PAR%, we included women with missing values for lifestyle factors (<2% over all questionnaires) in the high-risk category, to give the most conservative estimate.
We calculated the PAR% within 2 pre-specified subgroups. We examined a low-risk lifestyle among women with previous clinically diagnosed CHD and women without CHD. Diagnosis of CHD was reported every 2 years, and once a woman reported a diagnosis, she remained in the CHD group for the remainder of follow-up. Although smoking is a strong risk factor for SCD, more than 80% of women in the US are not current smokers.41 Therefore, we assessed the impact of the other lifestyle factors among nonsmokers. Analyses were conducted in SAS, version 9 (SAS Institute, Inc., Cary, NC). All p values are two sided, and a p-value of <0.05 was considered statistically significant.
Over 26 years of follow-up, we documented 321 cases of SCD among women of mean age 72±8 years. Baseline characteristics of the cohort according to the binary lifestyle score are presented in Table 1. Not smoking, exercise and healthy diet each were inversely associated with risk of SCD (Table 2) (P, linear trend <0.001). The association with BMI appeared J-shaped, with a nadir in SCD risk among women with a BMI of 21.0 -24.9 kg/m2 (P, nonlinear trend <0.001) (Table 2). When these risk factors were collapsed into low-risk binary categories, each lifestyle factor remained significantly associated with lower risk of SCD, even after controlling for the other low-risk factors (Table 3).
Overall, the binary lifestyle score was inversely associated with risk of SCD (p, deviation from linearity=0.55; p, linear trend<0.001) (Figure 1). The absolute risks of SCD were 22, 17, 18, 13 and 16 cases/100,000 person-years among women with 0, 1, 2, 3 and 4 low-risk factors, respectively. Women at low-risk for all 4 lifestyle factors (8% of the cohort) had a relative risk of SCD of 0.08 (95%CI: 0.03, 0.23) compared with women at low-risk for 0 factors (3% of the cohort). The expanded lifestyle score was also linearly associated with risk of SCD (p, deviation linearity = 0.12; p, linear trend <0.001). Women with an expanded lifestyle score ≥17 (15% of the cohort), compared to a score ≤8 (6% of the cohort), had a relative risk of SCD of 0.13 (95%CI: 0.07, 0.23).
The population attributable risk proportion for lack of adherence to each low-risk lifestyle factor is shown in Table 3. Although a strong risk factor for SCD, the estimated PAR% for smoking was only 11%, reflecting the low prevalence of current smoking (14%) in the cohort. For the other lifestyle factors, the PAR% ranged from 15% for BMI to 26% for exercise. Table 4 displays the PAR% associated with lack of adherence to an overall low-risk lifestyle among all women and in pre-specified subgroups. In the entire cohort, the PAR% associated with all four lifestyle factors was 81% (95%CI: 52%, 93%). Assuming causal relationships, these data suggest that 81% of SCD potentially may have been avoided had all women been in the low-risk group for all four lifestyle factors. Results were similar in sensitivity analyses that included unwitnessed deaths where the participant was documented to be symptom free within the preceding 24 hours and where circumstances suggested that the death could have been sudden (n=152). Among non-smokers, the PAR% associated with the remaining 3 lifestyle factors was 78% (46%, 92%). Finally, the PAR% associated with lack of adherence to a low-risk lifestyle was similar among women with and without clinically recognized CHD at the time of their most recent questionnaire (Table 4).
A low-risk lifestyle (not smoking, regular exercise, prudent diet and healthy weight) was linearly and inversely associated with risk of SCD among women. Women at low-risk for all 4 lifestyle factors had a 92% lower risk of SCD compared to women at low-risk for 0 lifestyle factors. If these associations are causal, hypothetically, 81% of SCD within this cohort may have been prevented if all women adhered to the low-risk lifestyle. Among women without diagnosed CHD, where the majority of SCDs occur, it is possible that 79% of SCD may be attributed to unhealthy lifestyle practices.
Primary prevention of SCD in women is of particular concern. Compared with men, women are 50% less likely to have severe left ventricular dysfunction and 66% less likely to be diagnosed with CHD before SCD, and therefore are less likely to meet current guideline recommendations for beta-blocker therapy or prophylactic ICD placement.42 Prevention efforts that can be applied across broader populations, such as healthy lifestyle practices, are crucial to prevent SCD, particularly among women.
Substantial evidence supports the benefit of lifestyle modification for the prevention of SCD. Smoking cessation has been associated with reductions in SCD risk,43 while regular physical activity has been inversely associated with lower risk of SCD in observational studies.8, 9 A Mediterranean-style diet was associated with lower risk of CVD in clinical trials44 and observational studies45. The association between the Mediterranean diet and CVD appears stronger for fatal, compared to nonfatal events31, which may be driven partially through protection against ventricular arrhythmias and SCD. Furthermore, several key components of a Mediterranean-style diet, including nuts, fish and omega-3 fats, and moderate intake of alcohol, have been associated with lower risk of SCD.10-14, 32, 33 Consistent with prior evidence, we found a strong inverse association between the aMed diet score and risk of SCD.
The J-shaped association between BMI and risk of SCD parallels the association with all-cause mortality.46 The elevated risk among women with BMI <21.0 is likely biased by reverse causation from preexisting disease, residual confounding by smoking, and effect modification by age.47, 48 BMI in midlife, which is largely unaffected by underlying disease, may quantify more accurately the effect of adiposity on SCD.49
More than 80% of the women in the US are not current smokers,41 however, the prevalence of other healthful habits is low.41 Among women 45-74 years old in the US, fewer than 40% maintain a BMI <25 kg/m2, 25% drink light-to-moderate amounts of alcohol and 22% exercise regularly at a light-to-moderate intensity.41 Nationally representative data are not available for the aMed score, but data from NHANES suggest that poor dietary habits are highly prevalent.28 Our data suggest that a substantial portion of SCD risk among nonsmokers was associated with poor diet, lack of exercise and unhealthy weight. Improvement in these lifestyle factors, while ultimately a personal choice, may be facilitated through changes in environmental settings and social norms, in part through public health policies that promote more healthful lifestyle choices.50
There are several limitations to our measure of the PAR% that warrant consideration. First, the PAR% assumes a casual relationship between the low-risk lifestyle factors and risk of SCD. Our study was not randomized, and therefore, this is a large assumption. However, a long-term trial assessing the effects of multiple lifestyle factors on risk of SCD, particularly for primary prevention, has inherent challenges including the necessity of a large sample size and long duration of follow-up and ensuring participant adherence to assigned dietary and exercise prescriptions, for example.51 In lieu of such data, carefully performed observational studies provide a reasonable approach for evaluating the association of multiple lifestyle factors on SCD risk.
Second, to estimate the PAR%, we dichotomized each lifestyle factor, although the relationships between the lifestyle factors and SCD more complex. When we used the expanded risk score, which accounts for the associations across the distribution of the lifestyle factor, the results were similar to the binary lifestyle score. Further, due to its distributive property, the PAR% from a multilevel exposure equals the PAR% calculated from collapsing the categories into a binary variable.39 The simplicity of binary cutpoints for the lifestyle factors mirrors the dichotomous cutpoints used to define low-risk for clinical risk factors, (i.e. total cholesterol <200 mg/dl or blood pressure<120/80) and may help provide discrete guidance for patients in the clinical setting. The set of binary low-risk factors that we utilized here for the prevention of SCD is also similar to an a priori low-risk lifestyle related to lower risk of CHD15, stroke16, CVD mortality21, diabetes,17 and cancer18. A single message for the prevention of CVD and other chronic diseases provides a simple strategy to minimize overall morbidity and premature death.
The PAR% is valid only when the relative risks and prevalence estimates used to calculate the PAR% are unbiased. The high degree of homogeneity in this cohort minimizes confounding by socioeconomic status and potentially other factors associated with a healthy lifestyle.15 We used multivariable models to adjust for additional confounders, however, the potential for residual confounding remains. Although measurement error in self-reported variables is unavoidable, information bias is minimal among these nurses who provide valid information on questionnaires.24, 25 Moreover, such error is likely to be non-differential with respect to SCD, and likely underestimate the true effect.
The PAR% is population-specific, thus the PAR% estimated among mainly Caucasian female health professionals may not be generalizable to men, or to women of other ethnicities. The prevalence of low-risk factors in the NHS is similar to the prevalence among US Caucasian women, but higher than the prevalence among black and Hispanic women.41 Additionally, incidence of SCD is greater, and survival after cardiac arrest is lower, among black Americans.52 Therefore, the impact of a low-risk lifestyle may be greater in more racially diverse populations.
We focused on the influence of modifiable lifestyle habits on SCD. It should be acknowledged that favorable levels of clinical risk factors, such as blood pressure and diabetes, are also associated with lower SCD risk.4 The association between lifestyle factors and SCD is at least partially mediated through these clinical risk factors; however, these later medical conditions are also influenced by factors other than lifestyle. Therefore, we did not include these clinical risk factors in our PAR estimate.
Our study has several important strengths. The repeated assessments of lifestyle factors allow us to update lifestyle habits throughout follow-up. The large number of rigorously confirmed SCDs, which is a difficult phenotype to classify in population studies, is a unique strength. Although we likely missed cases of SCD within this cohort, the high specificity of our defined cases provides a less biased risk estimate.53 Finally, we provide confidence intervals surrounding the PAR%, which are essential for describing estimation uncertainty, but not always presented.54
The primary prevention of SCD remains a major public health challenge because most SCD occurs among individuals not identified as high risk. In this cohort, adherence to an overall healthy lifestyle was associated with a lower risk of SCD among women and may be an effective strategy for the prevention of SCD. Because SCD accounts for over 50% of CHD mortality, widespread adoption of a healthy lifestyle in the population may make a substantial impact on reaching the American Heart Association's 2020 Impact Goal of further lowering CVD mortality.28
This study was supported research grants CA87969 and HL034594 from the National Institutes of Health and an Established Investigator Award from the American Heart Association to Dr. Albert. Dr Chiuve is supported by NIH grant K99 HL097068 and a Clinical Research Program Award from the American Heart Association. Dr. Chiuve and Dr. Albert had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
The authors report no conflicts of interest.
Author contributions: Design and conduct of the study: Stephanie Chiuve; Christine AlbertCollection of data: Meir Stampfer, JoAnn Manson, Christine Albert
Analysis of the data: Stephanie Chiuve
Interpretation of the data: Stephanie Chiuve, Teresa Fung, Kathryn Rexrode, Donna Spiegelman, JoAnn Manson, Meir Stampfer, Christine Albert
Preparation of the manuscript: Stephanie Chiuve
Review and approval of manuscript: Teresa Fung, Kathryn Rexrode, Donna Spiegelman, JoAnn Manson, Meir Stampfer, Christine Albert.
The sponsor played no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript