Chocolate consumption and risk of heart failure in the Physicians’ Health Study

Abstract

Aims

To test the hypothesis that chocolate consumption is associated with a lower risk of heart failure (HF).

Methods and Results

We prospectively studied 20,278 men from the Physicians’ Health Study. Chocolate consumption was assessed between 1999 and 2002 via a self-administered food frequency questionnaire and HF was ascertained through annual follow-up questionnaires with validation in a subsample. We used Cox regression to estimate multivariable adjusted relative risk of HF. During a mean follow-up of 9.3 years, there were 876 new cases of HF. The mean age at baseline was 66.4 ± 9.2 years. Hazard ratios (95% CI) for HF were 1.0 (ref), 0.86 (0.72–1.03), 0.80 (0.66–0.98), 0.92 (0.74–1.13), and 0.82 (0.63–1.07), for chocolate consumption of less than 1/month, 1–3/week, 2–4/week, and 5+/week, respectively, after adjusting for age, body mass index (BMI), smoking, alcohol, exercise, energy intake, and history of atrial fibrillation (p for quadratic trend = 0.62). In a secondary analysis, chocolate consumption was inversely associated with risk of HF in men whose BMI was <25 kg/m² (HR (95% CI) = 0.59 (0.37–0.94) for consumption of 5+ servings/week, p for linear trend = 0.03) but not in those with BMI of 25+ kg/m² (HR (95% CI) = 1.01 (0.73–1.39), p for linear trend = 0.42, p for interaction=0.17).

Conclusions

Our data suggest that moderate consumption of chocolate might be associated with a lower risk of HF in male physicians..

Introduction

Heart failure (HF) is a major comorbidity in older adults and remains a public health burden in the United States since it is associated with high healthcare expenditures.^1,2 At 40 years of age, the lifetime risk of HF is 1 in 5, and by age 65, HF incidence approaches 10 per 1000 individuals.^3,4 HF contributes to 1 in 9 deaths, and the costs of HF are estimated to be $31 billion annually and anticipated to reach $70 billion by 2030.⁴ Thus it is important to improve primary prevention of HF. Currently, lifestyle factors, such as regular exercise, and diet, including moderate alcohol intake, and consumption of fruits, vegetables, breakfast cereals, and fish high in omega-3 fatty acids, have been shown to be related to lower risk of HF.^5,6

Prior research has suggested that dark chocolate and cocoa intake are associated with reduced risk of cardiovascular disease and cardiovascular mortality.^7–10 The benefit of chocolate consumption on risk of CVD may be due to favorable effects of cocoa products on blood pressure, which is a major risk factor for HF.^10–13

One prior study assessed the impact of chocolate intake on the incidence of HF in the Swedish Mammography Cohort.¹⁴ In 9 years of follow-up, for women who consumed 1–3 servings/month, there was a 26% lower risk of HF compared to no regular intake (95% CI = 5%–42%), and a 32% lower risk of HF (95% CI=7%–50%) with consumption of 1–3 servings/week, and there was no association with 3 or more servings per week.¹⁴ Such data could have been confounded by overall healthy diet among chocolate consumers. Since no prior research has been published on the relation of chocolate intake with HF in healthy men we sought to assess whether moderate chocolate consumption is associated with lower risk of HF in US male physicians. In a secondary aim, we examined whether the relation between chocolate intake and risk of HF differs in lean versus overweight and obese individuals.

Methods

Study population

The Physicians’ Health Study (PHS) I is a completed, randomized, double-blind, placebo-controlled trial designed to study low-dose aspirin and β-carotene for the primary prevention of cardiovascular disease and cancer among US male physicians. In 1997, the PHS II enrolled 7,641 physicians from the PHS I and 7,000 new physicians to study the effects of β-carotene, vitamin C, E, and a multivitamin on cardiovascular disease and cancer. Detailed descriptions of the PHS I and II have previously been published.^15,16 Of the 29,071 total participants in the PHS, 21,075 completed a food frequency questionnaire (FFQ) between 1999 and 2002 and were alive at baseline. We excluded individuals with prevalent HF at the time of the FFQ (n=470) and missing data for chocolate consumption (n=327), for a final sample of 20,278 used in the present analysis. Each participant gave written informed consent, and the institution review board at Brigham and Women’s Hospital approved the study protocol.

Ascertainment of HF in the PHS

HF outcomes in the PHS were determined with the use of annual follow-up questionnaires mailed to each participant to obtain information on compliance with the intervention and the occurrence of new medical diagnoses. Cases of HF were identified by self-reported diagnosis. HF diagnoses in the PHS were previously validated by reviewing medical records in a subsample with a high positive predictive value (91%) in the PHS.^17,18

Characterization of chocolate consumption

Information on chocolate consumption was obtained by using a FFQ. Participants were asked to report their average consumption of 1 oz of chocolate during the past year. Possible answers were never or less than once per month; 1–3/month; 1/week; 2–4/week; 5–6/week; 1/day; 2–3/day; 4–5/day; and 6+/day. For current analyses, we collapsed adjacent categories to obtain stable estimate (never or <1/month; 1–3/month; 1/week, 2–4/week, and 5+/week) because only 5% of subjects indicated consumption of chocolate of 1 or more times a day. The validity and reproducibility of FFQs have been previously published.^19,20

Other variables

Information on demographic variables, body mass index (BMI), cigarette smoking, exercise, alcohol consumption, and history of diabetes, hypertension, and MI were collected at baseline. Information on incident co-morbidities, such as diabetes, hypertension, and MI, were collected through annual follow-up questionnaires as previously described.

Statistical Analysis

We used Cox regression analysis to calculated hazard ratios (HRs) and corresponding 95% confidence intervals (CIs). Person time was calculated from date of food frequency questionnaire to first date of HF, date of death, or date of last available follow-up. Proportional hazards assumption was assessed by including an interaction with logarithmic person time in the model. We used generalized additive models to assess whether fitting continuous variables as such or as categorical variables improved the fit of the model. We assessed confounding by age (<=55, >55–65, >65–75, >75), BMI (continuous), race (white/non-white), smoking (never, past, current), alcohol consumption (never, 1–3/month, 1–6/week, and 1+/day), exercise (<1/day, 1–2 days/week, 3–4 days/week, 5–7 days/week), caloric intake (tertiles), sodium intake (tertiles), fruit and vegetable intake (tertiles), red meat intake (tertiles), and history of atrial fibrillation (yes/no). A variable was considered a confounder if its adjustment led to a ≥ 10% change in the regression β coefficient for any chocolate category. A p value for linear trend was obtained by creating a new variable that was assigned median chocolate intake in each category, and p for quadratic trend was obtained by fitting a quadratic term for median chocolate consumption in a hierarchical model.

The initial model was adjusted for age. The parsimonious model accounted for age, BMI, smoking, alcohol consumption, exercise, caloric intake, and history of atrial fibrillation. We then considered hypertension (yes/no) and diabetes (yes/no) as potential mediators and did not adjust for them in the multivariable model. For each categorical variable we created indicator variables for missing observations.

In secondary analysis, we stratified BMI into <25 kg/m² (n=8,907) and ≥25 kg/m² (n=11,368). Analysis was completed using SAS version 9.3 (SAS Institute, Cary, North Carolina). All p values were 2-tailed, and the significance level was set at an alpha of 0.05.

Results

The characteristics of the 20,278 US male physicians are presented in Table 1 according to chocolate consumption. The mean age of study participants at baseline was 66.4 ± 9.2 (range 50.2–97.6). Median chocolate consumption was 1–3 time per month. Compared to consumption of <1/month, higher chocolate consumption was associated with being white, lower alcohol consumption, a higher proportion of never smokers, a lower prevalence of baseline diabetes, hypertension, MI, and atrial fibrillation, and higher energy intake.

Table 1.

Baseline characteristics of 20,278 participants in the Physicians’ Health Study according to chocolate intake

	Chocolate intake (servings/week)
Characteristics1	0	<1	1	2–4	5+	P for trend
n	5359	5907	3919	3234	1859
Age (y)	67.0 ± 8.8	66.0 ± 8.9	66.2 ± 9.2	66.2 ± 9.6	66.7 ± 10.1	0.39
BMI (kg/m2)	25.6 ± 3.3	25.9 ± 3.4	25.8 ± 3.2	25.8 ± 3.3	25.6 ±3.3	0.24
White	85.4	91.3	94.2	95.6	95.7	<0.01
Smoking status
Never	50.0	54.0	56.9	57.8	56.3	<0.01
Past	46.0	42.9	39.9	38.8	40.0	<0.01
Current	3.8	3.1	3.2	3.3	3.6	0.47
Alcohol consumption
Rarely/Never	17.0	15.4	16.8	18.6	20.3	<0.01
1–3×/month	6.6	7.3	8.0	8.8	8.7	<0.01
1–6×/week	35.6	38.7	39.4	39.7	37.7	<0.01
1 or more/day	40.3	38.0	35.1	32.3	32.8	<0.01
Exercise (days per week)
<1	38.0	37.3	35.8	36.5	37.6	0.20
1–2	15.3	16.7	17.3	16.3	15.3	0.56
3–4	28.1	29.2	30.2	30.9	29.1	0.03
5–7	16.7	15.3	15.1	14.4	16.3	0.08
Prevalent diabetes	11.4	6.5	5.2	5.0	4.8	<0.01
Prevalent hypertension	49.6	46.2	44.4	42.5	43.3	<0.01
Prevalent MI	4.8	3.9	3.0	3.0	3.7	<0.01
Prevalent atrial fibrillation	9.1	7.9	7.9	8.6	8.6	0.50
Calories (kcal)	1534.9 ± 481.1	1626.2 ± 485.5	1705.6 ± 499.8	1795.8 ± 509.5	2033.6 ± 579.9	<0.01
Sodium (mg)	1356.6 ± 528.2	1455.4 ± 532.5	1527.1 ± 546.3	1576.0 ± 541.3	1690.7 ± 607.4	<0.01
Fruits and Vegetables (s/d)	4.8 ± 2.9	4.8 ± 2.6	4.9 ± 2.6	4.9 ± 2.5	5.2 ± 3.0	<0.01
Red Meat (s/d)	0.6 ± 0.6	0.7 ± 0.6	0.7 ± 0.6	0.8 ± 0.6	0.9 ± 0.7	<0.01

¹Data are presented as means ± SD, median (IQR), or percentages. Few participants had missing data: BMI (n=3), race (n=52), smoking (n=14), alcohol (n=117), exercise (n=349), calories (n=894), sodium (n=894), fruit and vegetable intake (n=1), red meat intake (n=21)

In a Cox regression adjusted for age, HRs for HF were 1.00 (reference), 0.86 (0.72–1.02), 0.79 (0.65–0.96), 0.91 (0.74–1.11), and 0.82 (0.64–1.06) for chocolate consumption of <1/month, 1–3/month, 1/week, and 2–4/week, and 5+/week respectively (p for linear and quadratic trend = 0.35 and 0.51, respectively). In a parsimonious model adjusted for age, BMI, smoking, alcohol, exercise, energy intake, and prevalent atrial fibrillation, compared to consumption of <1/month, consumption of 1/week, was associated with a 20% lower risk of HF (95% CI: 2% to 34%), Table 2. As expected, additional adjustment for hypertension and diabetes slightly attenuated this relationship: HRs for HF were 1.00 (reference), 0.90 (0.76–1.08), 0.86 (0.70–1.05), 0.98 (0.80–1.21), and 0.89 (0.68–1.16) across categories of chocolate consumption.

Table 2.

Hazard ratios (95% CI) for heart failure according to chocolate consumption in the Physicians’ Health Study

			Hazard ratios (95% CI)
Frequency of chocolate intake	Cases/Person- time	Crude Incidence Rate (per 1000 person-years)	Age Adjusted	Model 11
Never or <1/month	259/48,669	5.32	1.00	1.00
1–3/month	243/55,622	4.37	0.86 (0.72–1.02)	0.86 (0.72–1.03)
1/week	153/37,119	4.12	0.79 (0.65–0.96)	0.80 (0.66–0.98)
2–4/week	144/30,585	4.71	0.91 (0.74–1.11)	0.92 (0.74–1.13)
5+/week	77/17,128	4.50	0.82 (0.64–1.06)	0.82 (0.63–1.07)
p for linear trend			0.35	0.41
p for quadratic trend			0.51	0.62

¹Adjusted for age (≤55, 55–≤65, 65–≤75, 75+), BMI (continuous), alcohol (never, 1–3/month, 1–6/week, and 1+/day), smoking (never, past, current), exercise (<1/day, 1–2 days/week, 3–4 days/week, 5–7 days/week), caloric intake (tertiles), and prevalent atrial fibrillation (yes/no)

In secondary analyses, we examined the relation of chocolate with risk of HF stratified by BMI, and showed a significant linear relation in individuals with a BMI <25 kg/m² (HRs for HF were 1.00 (reference), 0.86 (0.65–1.15), 0.70 (0.50–0.99), 0.72 (0.50–1.04), 0.59 (0.37–0.94), across consecutive categories of chocolate consumption (p for linear trend=0.03). For individuals with a BMI ≥25 kg/m², there was no significant association with HF (HRs and 95% CIs: 1.00 (reference), 0.87 (0.69–1.09), 0.88 (0.68–1.13), 1.06 (0.82–1.37), 1.01 (0.73–1.39) for chocolate consumption of <1/month 1–3/month, 1/week, and 2–4/week, and 5+/week respectively, p for linear trend=0.42, p for interaction (BMI*chocolate)=0.17, Figure 1.

Figure 1.

Hazard ratios (95% CI) for heart failure according to chocolate consumption stratified by body mass index1,2

1. Adjusted for age (≤55, 55–≤65, 65–≤75, 75+), BMI (continuous), alcohol (never, 1–3/month, 1–6/week, and 1+/day), smoking (never, past, current), exercise (<1/day, 1–2 days/week, 3–4 days/week, 5–7 days/week), caloric intake (tertiles), and prevalent atrial fibrillation (yes/no)

2. P for interaction = 0.17

Discussion

In this prospective cohort study, we found that moderate chocolate consumption was associated with a lower risk of HF in male physicians after adjustment for important confounders. In a secondary analysis, this inverse association was slightly stronger and stepwise in lean individuals.

Chocolate intake has been previously shown to be associated with lower blood pressure.^10–13 In short term trials, cocoa intake resulted in a significant reduction in mean systolic blood pressure by 2.77 mmHg (95% CI = −4.72, −0.82), and a reduction in diastolic blood pressure by 2.20 mmHg (95% CI = −3.46, −0.93).¹¹ Elevated blood pressure is a major risk factor for HF.^21–24 Our findings could be partly attributable to a decrease in blood pressure since we found that including hypertension in the model slightly attenuated our results.

Previous randomized control trials have suggested that flavonoids in chocolate may be responsible for the beneficial effect of chocolate on blood pressure, along with other risk factors for HF, including increasing HDL cholesterol, improving endothelial function, and reducing inflammation.^25,26 Cocoa products are rich is plant phytochemicals, especially flavonoids, with strong antioxidant properties.²⁷ Experiments with endothelial cell cultures have shown that flavonoids reduce the down-regulation of nitric oxide synthase, thereby promoting the production of nitric oxide and help maintain normal cardiac function.^28,29 By increasing nitric oxide production and decreasing oxidative stress, flavonoids could be playing a role to lower the risk of HF. In patients with HF, a randomized control trial found that feeding flavanol-rich chocolate significantly improved endothelial function suggesting an improvement in vascular function.³⁰

To our knowledge, only one prior study has prospectively tested whether chocolate consumption is related to risk of HF in healthy individuals.¹⁴ In the Swedish Mammography Cohort of 31,823 women, compared to no chocolate consumption, multivariable adjusted HFs for HF were 0.74 (0.58–0.95), 0.68 (0.50–0.93), 1.09 (0.74–1.62), and 1.23 (0.73–2.08) for chocolate intake of 1–3/month, 1–2/week, 3–6/week, and ≥ 1 serving/day, which is similar, albeit a stronger association, when compared to our current study. In individuals with a prior myocardial infarction in the Stockholm Heart Epidemiology Program, between 1/month and ≤1/week of chocolate consumption was associated with a 32% lower risk of HF (95% CI = 3%-53%) compared to no consumption.³¹ This suggests that chocolate consumption may be beneficial even in patients with prior coronary heart disease. The authors did not stratify by BMI for risk of HF, however, moderate chocolate consumption was associated with risk of cardiac mortality in patients with a BMI < 30 kg/m², but not in those with a higher BMI.³¹

In a secondary analysis, we found the association between chocolate consumption and heart failure in lean subjects to be stronger, although statistically not significantly different from overweight/obese males. This finding is worth pursuing in an independent cohort. It is possible that the benefits from flavonoid consumption in overweight individuals could be outweighed by the overall harm from metabolic syndrome. Alternatively, low risk of HF in people with normal BMI makes it easier to detect a small effect size compared to overweight and obese individuals.

Our study had a few limitations. Due to the observational design of the current study, we cannot exclude residual or unmeasured confounding as an alternate explanation of our results. Chocolate consumption could have been misclassified due to self-reports, because chocolate was only assessed at baseline, and because we didn’t have data on amounts of chocolate consumed. Additionally, our study did not have data on type of chocolate consumed, so it is unclear whether we would have obtained stronger results had we been able to restrict analysis to dark chocolate. We did not collect data to distinguish HF with and without preserved left ventricular function. Finally, our population was predominately Caucasian male physicians who may have had different behaviors than the general population; hence, our results may not be generalizable to women or other ethnic groups.

Strengths of our study include the prospective design of the PHS, a large sample size, the collection of a large number of key confounding variables, and a high positive predictive value (91%) of self-reported HF in male physicians.⁵

In conclusion, our data showed an inverse but non-linear relation between chocolate intake and risk of HF in male physicians.

Abbreviations

HF

heart failure

PHS

Physicians’ Health Study

FFQ

food frequency questionnaire

BMI

body mass index