OUP user menu

Chocolate consumption in relation to blood pressure and risk of cardiovascular disease in German adults

Brian Buijsse, Cornelia Weikert, Dagmar Drogan, Manuela Bergmann, Heiner Boeing
DOI: http://dx.doi.org/10.1093/eurheartj/ehq068 First published online: 30 March 2010

Abstract

Aims To investigate the association of chocolate consumption with measured blood pressure (BP) and the incidence of cardiovascular disease (CVD).

Methods and results Dietary intake, including chocolate, and BP were assessed at baseline (1994–98) in 19 357 participants (aged 35–65 years) free of myocardial infarction (MI) and stroke and not using antihypertensive medication of the Potsdam arm of the European Prospective Investigation into Cancer and Nutrition. Incident cases of MI (n = 166) and stroke (n = 136) were identified after a mean follow-up of ∼8 years. Mean systolic BP was 1.0 mmHg [95% confidence interval (CI) −1.6 to −0.4 mmHg] and mean diastolic BP 0.9 mmHg (95% CI −1.3 to −0.5 mmHg) lower in the top quartile compared with the bottom quartile of chocolate consumption. The relative risk of the combined outcome of MI and stroke for top vs. bottom quartiles was 0.61 (95% CI 0.44–0.87; P linear trend = 0.014). Baseline BP explained 12% of this lower risk (95% CI 3–36%). The inverse association was stronger for stroke than for MI.

Conclusion Chocolate consumption appears to lower CVD risk, in part through reducing BP. The inverse association may be stronger for stroke than for MI. Further research is needed, in particular randomized trials.

  • Chocolate
  • Cocoa
  • Blood pressure
  • Myocardial infarction
  • Stroke
  • Epidemiology

Introduction

Research on the health effects of cocoa and chocolate has accelerated during the last decade,1 in particular on cardiovascular health.2 Although chocolate is considered typically as a food people should indulge in only occasionally, several short-term experimental studies suggest that chocolate, already in amounts of several grams per day, improves endothelial37 and platelet function,6,8,9 and reduces blood pressure (BP)10 and markers of inflammation.11 Flavanols in cocoa are thought to be responsible for these effects.2

Despite this body of experimental evidence, observational studies on cocoa and the occurrence of cardiovascular diseases (CVD) have only sparingly been published. Two European cohort studies have suggested an inverse relation between the consumption of cocoa or chocolate and CVD.12,13 However, two studies in the USA showed absent14 or statistically weaker relations15 between chocolate consumption and CVD.

We investigated associations of habitual chocolate consumption with measured BP and incident myocardial infarction (MI) and stroke in middle-aged German individuals, sampled from the general population. Because chocolate appears to have a pronounced effect on BP, we hypothesized that chocolate consumption would relate stronger inversely with stroke than with MI.

Methods

Study population

The people in this study were participants in the Potsdam arm of the European Prospective Investigation into Cancer.16 In total, 27 548 participants from the general population of Potsdam and surroundings, 10 904 men mainly aged 40–65 years and 16 644 women mainly aged 35–65 years, took part in the baseline examination between 1994 and 1998.17 This examination included a food-frequency questionnaire, a BP measurement, anthropometric measurements, a personal interview that included questions about prevalent diseases, and a questionnaire on socio-demographic and lifestyle characteristics.

In three follow-up rounds, conducted about every 2–3 years, questionnaires sent by post were used to identify incident cases of chronic diseases. The response rates for the three rounds of follow-up were between 92 and 96%. For the current analysis, we also included questionnaires that were part of the fourth round of follow-up that were sent from September 2004 through December 2006; around 90% of the questionnaires were returned to the study centre. Informed consent was obtained from all participants. Study procedures were approved by the Ethics Committee of the Medical Association of the State of Brandenburg.

Exclusion criteria

Criteria for exclusion were prevalent CVD, use of antihypertensive medication, and missing BP data, amongst others (Figure 1). After exclusions, 19 357 participants (70.3%) were eligible for analysis. Participants who were excluded were on average older, more likely to be men, had less often a university degree, had a higher mean body mass index (BMI), and were more likely to have diabetes (P for all <0.0001). Mean chocolate consumption did not differ between participants who were included and excluded (6.1 vs. 6.3 g/day, P = 0.10).

Figure 1

Flow sheet indicating reason and number of excluded participants.

Baseline examination

Assessment of diet and chocolate consumption

Usual food intake in the year before baseline was assessed by a self-administered 148-item food-frequency questionnaire. Details about its validation and reproducibility have been reported previously.1821 For each item the frequency of consumption was asked in 10 categories, ranging from ‘never’ to ‘five times a day or more’. Pictures and portion sizes were given to estimate the quantities that were consumed. The consumed amount of each food item was calculated in grams per day. Chocolate consumption was asked by how frequent a chocolate bar of 50 g was consumed. Participants could indicate whether they consumed half, one, two, or three bars of chocolate.

In a subset of participants (n = 1568), dietary intake was also assessed by a single 24 h dietary recall. This information was used to estimate the amount of chocolate consumption across categories of chocolate consumption based on the food-frequency questionnaire and to identify the types of chocolate that were consumed.

Blood pressure measurement

Systolic and diastolic BP at baseline were measured in sitting position at the right arm at heart level.22 After the participants rested for 15–30 min, three consecutive readings were performed with 2 min intervals using 11 automatic oscillometric devices (BOSO-Oscillomat, Bosch & Sohn, Jungingen, Germany). The reproducibility of these devices was monitored regularly and the validity was tested in a subset of the population by using aneroid devices; only very small systematic differences were detected between both methods.23 Because the mean of the second and third readings has been shown to be the most reliable estimate of a person's BP in the EPIC-Potsdam study,22 this variable was used in the analysis.

Anthropometrics and other covariates

Body weight, height, and waist circumference were measured by trained staff according to standard procedures with participants wearing light underwear.24 Body mass index was calculated by dividing weight (kg) by height (m) squared. Basal metabolic rate was calculated from age, sex, weight, and height.25 Information on smoking habits, education, and occupational and leisure time physical activity were obtained by questionnaire.

Case ascertainment

Possible cases of MI or stroke were identified by self-reports or death certificates. Self-reported information was obtained by at least one of the four follow-up questionnaires, which contained questions about physician-diagnosed CVD and the use of medication. Participants who reported post-baseline dietary changes due to CVD were also checked for incidence of CVD. To increase the sensitivity for incident stroke, follow-up questionnaires included additional questions about symptoms typical for stroke.26 All possible cases were verified by reviewing medical records from the hospital, by contacting the patients’ physician, or by review of the death certificate. These sources also provided the date of CVD incidence. The verification was done by two physicians working at the study centre (C.W. and Wolfgang Fleischhauer). Death within 28 days after a diagnosis of MI or stroke and an underlying cause of death from MI or stroke was considered as a fatal CVD case. All verified incident CVD cases were coded using the International Classification of Diseases (10th revision, ICD-10).27 Incident MI was defined as ICD-10 I21, incident stroke as ICD-10 I63.0–I63.9 (ischaemic stroke) or ICD-10 I60.0–I60.9 or I61.0–I61.9 (hemorrhagic stroke). Only confirmed incident CVD cases were included in the analysis.

Statistical analysis

Chocolate intake was adjusted for total energy intake by regressing its intake on total energy.28 Participants were categorized into quartiles based on the obtained residuals. Because our findings on BP and incident CVD did not differ between men and women, we combined them in all analyses.

Adjusted least-square means of systolic and diastolic BP and their corresponding 95% confidence intervals (CIs) were computed for quartiles of chocolate intake. Cox regression models were used to estimate relative risks (RRs) and 95% CIs of incident CVD. Entry-time was the participant's age at recruitment and exit-time was the age at either the diagnosis of MI or stroke, death, or return of the last follow-up questionnaire, whichever came first. All Cox models were stratified on baseline age. Tests for a linear trend across quartiles were performed by modelling the median value of chocolate consumption in each quartile as a continuous variable.

We also studied energy-adjusted chocolate intake as a continuous variable. Because chocolate consumption was skewed to the right (skewness = 6.6), we log-transformed the intake values. A Log2 transformation was applied, which, apart from its normalizing effects (skewness was reduced to −0.1), also is easy to interpret because it corresponds to effect measures for doubling chocolate intake. Transformed values were adjusted for energy intake by regressing them on Log2-transformed total energy intake. Using this variable, a competing risk analysis was performed in which separate regression coefficients for stroke and MI were compared using a Wald test.29

All analyses were adjusted according to three models. Model A included adjustments for age and sex. In model B, we adjusted additionally for alcohol intake (grams of ethanol per day), employment status (dummy variables for part-time, hourly worker, [pre-]retirement, and unemployed), BMI, waist circumference, smoking (dummy variables for former smoking, current smoking <20 cigarettes/day, and current smoking 20 cigarettes/day or more), occupational physical activity (dummy variables for medium and high activity), sports (hours per week), cycling (hours per week), education (dummy variables for technical school and university degree), and total energy intake. Finally, in model C, we also adjusted for the prevalence of diabetes and for energy-adjusted intakes of fruit, vegetables, red meat, processed meat, dairy, coffee, tea, and cereal fibre.

The association between chocolate intake and incident CVD was also assessed by using restricted cubic spline regression,30 with knots placed at the 5th, 50th, and 95th percentile of energy-adjusted chocolate intake (continuous variable). To limit the influence of extreme high chocolate intakes, participants in the top 2.5 percentile were omitted for this analysis.

Because of the possibility that chocolate lowers CVD risk by lowering BP, we estimated this effect by calculating the percentage of risk reduction that was explained by baseline BP.31 For this purpose two multivariate models were ran, one with and one without systolic and diastolic BP as continuous variables. In these models chocolate was analysed continuously after Log2-transformation and the proportional difference between the two regression coefficients was calculated along with 95% CIs.

Effect modification by sex and BMI (continuous) was evaluated by entering product terms along with main effects using the Log2-transformed chocolate variable. We repeated the analysis after excluding prevalent diabetes cases. Statistical tests were 2-sided and P-values <0.05 were used to indicate statistical significance. SAS version 9.2 (SAS Institute, NC, USA) was used for all analyses.

Results

Baseline characteristics

At baseline 92.3% of the study population reported to consume chocolate in the food-frequency questionnaire. Using data from a 24 h dietary recall method, conducted in 8.1% of the study population, mean chocolate intake was ∼6 g/day higher in the top quartile than in the bottom quartile (P linear trend adjusted for age and sex <0.0001). With increasing quartiles of chocolate consumption, the proportion of women increased and the mean intake of alcohol decreased (Table 1). Consumption of fruits, vegetables, and dairy was inversely related to chocolate consumption. Although chocolate intake was statistically significantly related with some other characteristics, the magnitude of these relations was weak or visually inconsistent across quartiles.

View this table:
Table 1

Selected baseline characteristics by quartiles of chocolate consumption

Quartiles of energy-adjusted chocolate consumptionP linear trenda
1234
Number4839483948394840
Chocolate intake, g/dayb1.71.93.37.5<0.0001
Age, year48.949.849.248.4<0.0001
Women, %41.567.174.964.0<0.0001
University degree, %36.840.440.138.60.13
Body mass index, kg/m225.625.725.825.70.18
Waist circumference (men/women), cm93.8/79.194.1/78.893.7/79.293.0/79.40.01/0.15
Smoking, %0.04
 Never47.548.945.946.9
 Former30.531.231.830.2
 Current <20 cigarettes/day15.714.316.316.4
 Current ≥20 cigarettes/day6.35.66.06.5
Ethanol, g/day19131212<0.0001
Sports activities, h/week1.021.011.031.000.65
Cycling, h/week1.951.731.781.840.14
Occupational activity, %<0.0001
 Light56.161.562.760.3
 Moderate35.432.631.533.1
 Heavy8.55.85.86.6
Prevalent diabetes, %3.02.93.13.00.71
Total energy intake, MJ/day10.68.27.79.0<0.0001
Energy-adjusted intake, g/day
 Fruit171155152154<0.0001
 Vegetables137131128124<0.0001
 Red meat283030280.09
 Processed meat70717265<0.0001
 Dairy247229217225<0.0001
 Coffee4334414384360.73
 Tea1271161191120.006
 Cereal fibre12121211<0.0001
  • Data given were for 19 357 men and women without a history of myocardial infarction and stroke, and not using antihypertensive medication.

  • All values other than the number of participants, sex, and age were adjusted for age and sex, including percentages. Waist circumference was adjusted for age only, age was adjusted for sex only, and sex was unadjusted.

  • aP for linear trend across quartiles based on ANCOVA.

  • bMean values based on a 24 h recall conducted in 1568 of the participants, adjusted for age and sex.

Because our food-frequency questionnaire asked for chocolate consumption in general, we used data of the 24 h dietary recall method to evaluate what kinds of chocolate were consumed most frequently. Among those who reported to consume chocolate at the day of the recall, 57% consumed milk chocolate, 24% consumed dark chocolate, and only 2% consumed white chocolate (of 17% the type of chocolate was unspecified). The consumption of dark chocolate (P = 0.001) and milk chocolate (P < 0.0001) increased across quartiles of chocolate consumption, whereas the consumption of white chocolate did not differ (P = 0.61).

Cross-sectional association between chocolate consumption and blood pressure

Chocolate consumption was related to a lower systolic and diastolic BP in a linear fashion (Table 2). After adjustment for age and sex, lifestyle variables, indicators of socio-economic status, dietary factors, and the prevalence of diabetes, the difference between top and bottom quartiles was 1.0 mm Hg for systolic BP (95% CI, −1.6 to −0.4 mm Hg; P linear trend = 0.0008) and 0.9 mm Hg for diastolic BP (95% CI, −1.3 to −0.5 mm Hg; P linear trend <0.0001). Similar findings were obtained after chocolate intake was studied as a continuous Log2-transformed variable. Findings were essentially similar after excluding participants with prevalent diabetes at baseline.

View this table:
Table 2

Cross-sectional association between chocolate consumption and blood pressure

Mean blood pressureMean blood pressure values (95% CI) for quartiles of energy-adjusted chocolate consumptionP linear trendaBeta for chocolate as a continuous variable (95% CI)b
1234
Number of participants4839483948394840
Mean systolic BP
 Model Ac127.0 (126.6–127.5)126.7 (126.3–127.1)126.4 (125.9–126.8)125.7 (125.3–126.1)<0.0001−0.29 (−0.39 to −0.19)
 Model Bd126.9 (126.5–127.4)126.6 (126.2–127.0)126.3 (125.9–126.7)126.0 (125.6–126.4)0.001−0.21 (−0.31 to −0.11)
 Model Ce126.9 (126.5–127.4)126.7 (126.3–127.1)126.3 (125.9–126.7)125.9 (125.5–126.3)0.0008−0.22 (−0.32 to −0.12)
Mean diastolic BP
 Model Ac82.8 (82.6–83.1)82.7 (82.4–83.0)82.8 (82.5–83.1)81.9 (81.6–82.1)<0.0001−0.20 (−0.26 to −0.13)
 Model Bd82.9 (82.6–83.2)82.6 (82.3–82.8)82.6 (82.4–82.9)82.1 (81.8–82.3)<0.0001−0.14 (−0.21 to −0.08)
 Model Ce82.9 (82.6–83.2)82.6 (82.3–82.9)82.6 (82.4–82.9)82.0 (81.8–82.3)<0.0001−0.16 (−0.22 to −0.09)
  • Data given were for 19 357 men and women without a history of myocardial infarction and stroke, and not using antihypertensive medication.

  • aOn the basis of modelling the median value of chocolate consumption in each quartile as a continuous variable.

  • bRegression coefficient in which chocolate intake was modelled as a continuous Log2-transformed variable.

  • cDerived from least-square-mean regression models adjusted for age and sex.

  • dAs model A with additional adjustment for alcohol intake, employment status, BMI, waist circumference, smoking status, occupational physical activity, sports, cycling, education, and total energy intake.

  • eAs model B with additional adjustment for energy-adjusted intakes of fruit, vegetables, red meat, processed meat, dairy, coffee, tea, and cereal fibre, and prevalence of diabetes.

Prospective association between chocolate consumption and incident cardiovascular disease

After a mean follow-up of 8.1 years (median, 8.3 years; interquartile range, 7.5–9.0 years), comprising roughly 155 000 person-years, 166 cases (n = 24 fatal) of MI and 136 cases (n = 12 fatal) of stroke had occurred. Eighty-four percent of strokes were of ischaemic nature.

Chocolate consumption analysed in quartiles or as a continuous variable after Log2 transformation was inversely related with CVD (Table 3). After controlling for age, sex, lifestyle factors, anthropometrics, dietary factors, and for the prevalence of diabetes (model C), the RR was 0.61 (0.44–0.87) for the combined end point of MI and stroke comparing top vs. bottom quartiles (P linear trend = 0.014). The RR of fatal CVD for chocolate modelled as a Log2-transformed variable was 0.86 (95% CI 0.73–1.00) after the same adjustments.

View this table:
Table 3

Prospective association between chocolate consumption and incident cardiovascular disease

OutcomeQuartiles of energy-adjusted chocolate consumptionP linear trendaRelative risk for chocolate as a continuous variable (95% CI)b
1234
Myocardial infarction and stroke combined
 Number of cases106647656
 Incident rate per 100 000 person-years272164195144
 Relative risk (95% CI)
  Model AcReferent0.75 (0.55–1.03)0.98 (0.72–1.03)0.68 (0.49–0.95)0.0490.92 (0.87–0.98)
  Model BdReferent0.70 (0.50–0.99)0.91 (0.65–1.27)0.63 (0.45–0.88)0.0200.92 (0.87–0.97)
  Model CeReferent0.71 (0.51–1.00)0.91 (0.65–1.27)0.61 (0.44–0.87)0.0140.92 (0.87–0.97)
Myocardial infarction
 Number of cases59304334
 Incident rate per 100 000 person-years1517711087
 Relative risk (95% CI)
  Model AcReferent0.69 (0.44–1.08)1.11 (0.74–1.68)0.82 (0.54–1.26)0.590.95 (0.88–1.03)
  Model BdReferent0.63 (0.39–1.02)1.00 (0.64–1.56)0.74 (0.48–1.15)0.350.94 (0.88–1.02)
  Model CeReferent0.65 (0.40–1.05)1.02 (0.65–1.60)0.73 (0.47–1.15)0.330.95 (0.88–1.02)
Stroke
 Number of cases47343322
 Incident rate per 100 000 person-years121878557
 Relative risk (95% CI)
  Model AcReferent0.80 (0.51–1.26)0.84 (0.53–1.34)0.54 (0.32–0.90)0.0210.89 (0.82–0.97)
  Model BdReferent0.77 (0.47–1.26)0.80 (0.48–1.33)0.52 (0.31–0.89)0.0220.90 (0.83–0.97)
  Model CeReferent0.79 (0.48–1.29)0.80 (0.48–1.35)0.52 (0.30–0.89)0.0220.90 (0.83–0.98)
  • Data given were for 19 357 men and women without a history of myocardial infarction and stroke, and not using antihypertensive medication.

  • aOn the basis of modelling the median value of chocolate consumption in each quartile as a continuous variable.

  • bRelative risk in which chocolate intake was modelled as a continuous Log2-transformed variable.

  • cDerived from Cox proportional-hazards regression, with age as underlying time variable, stratified by age at baseline, and adjusted for sex.

  • dAs model A with additional adjustment for alcohol intake, employment status, BMI, waist circumference, smoking status, occupational physical activity, sports, cycling, education, and total energy intake.

  • eAs model B with additional adjustment for energy-adjusted intakes of fruit, vegetables, red meat, processed meat, dairy, coffee, tea, and cereal fibre, and prevalence of diabetes.

The inverse relation was stronger for stroke than for MI. The adjusted RR (model C) for comparing top vs. bottom quartiles was 0.73 (95% CI 0.47–1.15) for MI and 0.52 (0.30–0.89) for stroke. In a competing-risk analysis, the Wald test for assessing the difference in association between chocolate and stroke vs. chocolate and MI yielded a P-value of 0.08.

The linearity of the shape of the associations was assessed with restricted cubic spline regression and jointly testing the cubic and quadratic terms for statistical significance (Figure 2). We found no departure from linearity regardless whether the outcome was MI, stroke, or both combined (for all P for non-linearity ≥0.25).

Figure 2

Relative risk for the association between chocolate consumption and incident major cardiovascular disease. The solid line indicates the relative risk and the dashed lines the 95% confidence intervals as obtained by restricted cubic spline regression with knots placed at the 5th, 50th, and 95th percentile of the distribution of energy-adjusted chocolate intake. The reference was set at the median intake of ∼10 g/day. Adjustments were made according to the covariates in model C.

There was no evidence of a differential association of chocolate intake with MI or stroke by sex or BMI (for all P for interaction >0.20). We repeated the analysis after excluding participants with prevalent diabetes, restricting the study sample to 18 775 participants. The RR (95% CI) per unit increase of log2-transformed chocolate intake was 0.92 (0.85–1.00) for MI and 0.91 (0.83–0.99) for stroke (adjusted according model C).

We evaluated whether the lower BP that was related to chocolate intake at baseline acted as a mediator in lowering CVD risk. Entering systolic and diastolic BP into the multivariate model C weakened the relation. Baseline BP explained 12% of the inverse relation between chocolate and the combined outcome of MI and stroke (95% CI 3–36%). These estimates were 16% for MI (95% CI −1 to 26%) and 10% for stroke (95% CI 1–47%).

To assess whether the inverse relation was specific for chocolate, we also analysed the intake of non-chocolate sugar confectionary. RRs (95% CIs) of the combined outcome of MI and stroke for increasing quartiles were 1.00, 0.89 (0.64–1.23), 0.71 (0.49–1.03), and 0.91 (0.66–1.27; P linear trend = 0.36; model C). Similar results were obtained when MI and stroke were analysed separately.

Discussion

In this cohort of German adults, consumption of 6 g of chocolate per day was associated with a 39% lower risk of the combined outcome of MI and stroke. This was partly (for 12%) explained by baseline BP, to which chocolate consumption was also inversely related. The inverse relation of chocolate consumption appeared stronger for stroke than for MI. Finally, the inverse relations of chocolate with BP and incident CVD were observed despite lower intakes of fruit and vegetables in people consuming more chocolate.

The BP-lowering effect of chocolate has been assessed in a number of intervention studies.10 Most of them used amounts of chocolate (∼100 g/day) that are unrealistic to consume each day. One experimental study, however, used only ∼6 g of dark chocolate per day for 18 weeks long, after which systolic BP was reduced with on average 2.9 mmHg and diastolic BP with on average 1.9 mmHg.32 In the Zutphen Elderly Study, men with usual intakes of 4 g of cocoa per day (equal to about 10 g of dark chocolate) had a 3.7 mmHg lower systolic BP and a 2.1 mmHg lower diastolic BP than non-consumers of cocoa.12 The current study, however, suggests that an increase in chocolate consumption with 6 g per day relates to a lower systolic and diastolic BP of on average only 1 mmHg. These weaker estimates may be explained by the fact that BP values were lower in our study compared with other studies. The preference of milk chocolate over dark chocolate by our participants may also have contributed to the weaker estimates because milk chocolate contains less cocoa and therefore less flavanols than dark chocolate.

The inverse relation between chocolate consumption and incident CVD appeared stronger for stroke than for MI in our study, even though a formal statistical test to check whether the association between MI and stroke differed did not reach the customary level of statistical significance (P = 0.08). This finding, however, does not stand alone. In the Women's Health Study the association between chocolate consumption and CVD also appeared stronger for stroke (RR consumers vs. non-consumers = 0.85, 95% CI 0.70–1.03) than for coronary disease (RR = 0.97, 95% CI 0.86–1.09).15 In the Nurses’ Health Study no association was observed between chocolate consumption and coronary disease.14 Even though one other study showed a strong inverse relation between chocolate consumption and cardiac death in coronary patients,13 it may be that consumption of chocolate relates stronger inversely to stroke than to coronary disease. Interestingly, two small experimental studies reported that cocoa consumption increases cerebral blood flow in humans.33,34 Whether this also translates to a lower risk of stroke needs to be shown. Further research in this area is needed before firm conclusions can be drawn.

Only 10–16% of the inverse relation between chocolate intake and CVD risk was explained by measured BP in our study. This may be an underestimation because we measured BP only once at baseline. Chocolate, however, improves also other cardiovascular risk factors, including endothelial37 and platelet function.6,8,9 These may also contribute to reducing CVD risk and offer alternative explanations for the lower risk of CVD in our study.

One limitation of our study was that chocolate consumption was estimated with only one item in our food-frequency questionnaire. This may have led to an underestimation of chocolate intake because chocolate and cocoa are added to other foods. Also, we could not distinguish between milk chocolate and dark chocolate, which is important because the cocoa content is lower in milk chocolate than in dark chocolate and because milk may affect the bioavailability of flavanols.35 At group level, we were able to identify the kinds of chocolate that were consumed using dietary data of a 24-recall recall, which was conducted in a subset of participants. On the basis of this information, milk chocolate was the most frequently consumed chocolate, followed by dark chocolate; only few participants consumed white chocolate.

Another limitation is that dietary intake, risk factors, and BP were assessed at baseline only; therefore, we assume that these variables remained stable over time. Also, although we adjusted for a wide range of possible confounding factors, it remains possible that our findings are explained by residual confounding. Finally, because possible incident CVD cases were mostly identified by self-report, we can not exclude the possibility of having false-negative cases. This is unlikely, however, to affect the RR, assuming that their number is balanced across the chocolate quartiles. Because we used only validated incident CVD cases, the potential of having false-positives was minimized. Other strengths of our study are the prospective design of the analysis on incident CVD and the validated BP measurements. Finally, because chocolate consumption was assessed before the first scientific studies on the health effects of cocoa were published, our study participants were unaware of any health benefits that chocolate may have and were therefore unlikely to modify their intakes accordingly.

In summary, our data show that consumption of low amounts of chocolate relates to a lower risk of CVD, which is partly explained by a lower BP. This inverse association appears to be stronger for stroke than for MI. Given these and other promising health effects of cocoa, it is tempting to indulge more in chocolate. Small amounts of chocolate, however, may become part of a diet aimed to prevent CVD only after confirmation by other observational studies and particularly by randomized trials.

Funding

The recruitment part of the EPIC-Potsdam Study was supported by the German Federal Ministry of Science (01 EA 9401) and the European Union (SOC 95201408 05F02). The follow-up of the EPIC-Potsdam Study was supported by the German Cancer Aid (70-2488-Ha I) and the European Community (SOC 98200769 05F02).

Conflict of interest: none declared.

Acknowledgements

We thank Wolfgang Fleischhauer for case ascertainment and Ellen Kohlsdorf for data management.

References

View Abstract