Professional Documents
Culture Documents
Larsson, SC 2016
Larsson, SC 2016
com
ORIGINAL ARTICLE
specified). The FFQ had eight predefined frequency consump- Tests for linear trend across categories of chocolate consump-
tion categories (never, 1–3/month, 1–2/week, 3–4/week, 5–6/ tion were performed by creating a variable containing the
week, 1/day, 2/day and ≥3/day). Participants were categorised median value of chocolate consumption for each category and
into four groups according to their chocolate consumption: treating this variable as a continuous variable in the Cox model.
never, 1–3/month, 1–2/week and ≥3–4/week. The five highest In a sensitivity analysis, we excluded cases diagnosed during the
frequency categories were collapsed because only 8.7% of parti- first 3 years of follow-up. We examined whether the association
cipants consumed chocolate ≥3–4 times/week. The validity and between chocolate consumption and MI risk was modified by
reproducibility of the FFQ have been previously published.12 sex, high educational level (high school or university), over-
weight (BMI ≥25 kg/m2) and history of diabetes, hypertension
Assessment of covariates and hypercholesterolaemia at baseline. The statistical signifi-
Information on education, family history of MI before 60 years cance of the potential effect modifications was tested using the
of age, smoking, aspirin use, physical activity (including walking/ likelihood ratio test, comparing models with and without inter-
bicycling and leisure-time exercise), body weight, height, history action terms. Analyses were performed with SAS V.9.3 (SAS
of diabetes, hypertension and hypercholesterolaemia, and Institute, Cary, North Carolina, USA). All p values were 2-tailed
alcohol intake was obtained by a self-administered questionnaire. and the significance level was set at an α of 0.05.
Information on history of diabetes and hypertension was also
obtained from the Swedish National Diabetes Register (diabetes Meta-analysis
only) and the Swedish National Patient Register. Body mass We performed a meta-analysis that incorporated results from the
index (BMI) was calculated as weight (in kg) divided by the current Swedish study into findings from previous prospective
square of height (in metres). Pack-years of smoking history were studies of the association of chocolate consumption with risk of
computed by multiplying the number of packs of cigarettes MI or IHD. Inclusion criteria were prospective study and results
smoked per day by the number of years of smoking. for chocolate intake in relation to MI/IHD incidence or mortal-
ity. We excluded letters, abstracts, systematic reviews and
Case ascertainment meta-analyses; case–control, cross-sectional, ecological and
Incident MI cases (including fatal cases) were ascertained by experimental studies. We searched PubMed (http://www.ncbi.
linkage of participants (using the unique personal registration nlm.nih.gov/pubmed) and EMBASE (http://www.embase.com)
number assigned to each Swedish resident) to the Swedish from inception until 4 February 2016, using the search terms
National Patient Register and the Swedish Cause of Death chocolate or cocoa combined with cardiovascular disease or
Register at the National Board of Health and Welfare. MI was myocardial infarction or heart disease. No restrictions were
classified using the International Classification of Diseases 10th applied. Reference lists of pertinent articles were reviewed to
Revision code I21 for acute MI. Deaths were ascertained by identify further relevant studies. From each study, we extracted
linkage with the Swedish Cause of Death Register. the following data: the first author’s last name, publication year,
the name of the study, study location, sex and age of partici-
Statistical analysis pants, years of follow-up, sample size (number of events and
For each participant, person-years of follow-up accrued from 1 total number of participants), chocolate intake categories, cov-
January 1998 until the date of diagnosis of MI, date of death ariates adjusted for in the multivariable model and the most
(from any cause) or 31 December 2010, whichever came first. fully adjusted RRs.
Cox proportional hazards regression models were used to esti- We combined the RRs for the highest versus lowest category
mate hazard ratios (hereafter referred to as relative risks (RR)) of chocolate consumption. We also performed a dose–response
with 95% CI of MI according to categories of chocolate con- meta-analysis using the method by Greenland and Longnecker13
sumption. The first multivariable model was stratified by base- and Orsini et al14 to compute the trend from the correlated log
line age (in years) and sex and included education (less than RRs across categories of chocolate consumption. Studies were
high school, high school, university), family history of MI included in the dose–response meta-analysis if they reported
before 60 years of age (no, yes), smoking (never; past <20 or RRs for at least three categories of chocolate consumption. If
≥20 pack-years; current <20 or ≥20 pack-years), aspirin use results were reported in servings of chocolate, we assumed that
(never, 1–6 tablets/week, ≥7 tablets/week or users with one serving equals 30 g chocolate, which is the approximate
unknown number of tablets), walking/bicycling (almost never, serving size of chocolate in Swedish men.12 Study-specific RRs
<20 min/day, 20–40 min/day, 40–60 min/day, >1 h/day), exer- were combined using a random-effects model,15 which consid-
cise (<1 h/week, 1 h/week, 2–3 h/week, 4–5 h/week, >5 h/ ers both within-study and between-study variation (weighting
week) and intakes of total energy (kcal/day; continuous), was based on the inverse of the variance). The I2 statistic16 was
alcohol (g/day; quintiles), processed meat (servings/week; quin- used to quantify heterogeneity between studies. Publication bias
tiles) and fruits and vegetables (servings/day; quintiles). In a was assessed with Egger’s test.17 We used Stata V.14.1
second multivariable model, we included the above covariates (StataCorp, College Station, Texas, USA) to analyse the data.
plus potential intermediates: BMI (kg/m2; continuous) and
history and diagnosis of diabetes (no, yes), hypertension (no, RESULTS
yes) and hypercholesterolaemia (no, yes) at baseline. A separate Swedish study
category for missing was used to handle missing covariate data Over 13 years (mean 12.0±2.6 years, median 13 years) of
(missing was ≤3%, with the exception for exercise (5.5%)). follow-up, we identified 4417 incident cases of MI (3067 in
Further adjustment for intake of unprocessed red meat, fish, men and 1350 in women), including 897 fatal cases, among the
dairy products, sweetened beverages, coffee and tea did not 67 640 Swedish adults. Compared with non-consumers of choc-
alter the results; therefore, these variables were not included in olate, men and women who consumed ≥3–4 servings/week of
the multivariable model. The proportional hazards assumption chocolate were more likely to have a university education but
was tested using Schoenfeld residuals and was found to be less likely to be current smokers, overweight and to have a
satisfactory. history of diabetes, hypertension and hypercholesterolaemia
1018 Larsson SC, et al. Heart 2016;102:1017–1022. doi:10.1136/heartjnl-2015-309203
Downloaded from http://heart.bmj.com/ on December 20, 2016 - Published by group.bmj.com
Table 1 Baseline age-standardised characteristics by categories of chocolate consumption in 67 640 Swedish men and women
Category of chocolate consumption, servings
Characteristic* 0 (n=10 037) 1–3/month (n=34 231) 1–2/week (n=17 487) ≥3–4/week (n=5885) p for trend†
Age, years 61.0 (9.2) 59.3 (8.9) 58.8 (9.0) 60.7 (9.6) <0.001
Sex, men, % 53 52 57 57 <0.001
Postsecondary education, % 17 18 22 24 <0.001
Family history of myocardial infarction, % 16 15 15 14 <0.001
Current smoker, % 26 24 23 26 0.83
Aspirin use ≥7 tablets/week, % 6.5 6.5 6.0 7.2 0.48
Walk/bicycle ≥40 min/day, % 36 34 32 34 <0.001
Exercise ≥2 h/week, % 58 58 57 55 <0.001
Body mass index ≥25 kg/m2, % 53 50 48 44 <0.001
Diabetes, % 14 3.8 3.1 3.2 <0.001
Hypertension, % 24 20 19 18 <0.001
Hypercholesterolaemia, % 12 10 10 10 0.002
Total energy intake, kcal/day 2100 (830) 2200 (800) 2400 (850) 2700 (940) <0.001
Alcohol intake, g/day 11 (22) 10 (16) 11 (15) 12 (17) <0.001
Processed meat, servings/week 4.8 (4.6) 4.8 (3.9) 5.4 (4.1) 5.9 (4.9) <0.001
Fruits and vegetables, servings/day 4.6 (3.0) 4.4 (2.5) 4.4 (2.4) 4.6 (2.7) 0.37
*Data represent mean (standard deviation) or percentage.
†Assessed by using linear regression for continuous variables and χ² test for categorical variables.
(table 1). They also had higher intakes of total energy and pro- with diabetes (multivariable model 2: RR 0.69, 95% CI 0.46 to
cessed meat. 1.05, for highest vs lowest category) than in non-diabetics (cor-
Chocolate consumption was inversely associated with MI risk responding RR 0.93, 95% CI 0.82 to 1.06) ( p for interaction
(table 2). In the minimally adjusted multivariable Cox model, between chocolate consumption and diabetes in relation to MI
men and women who consumed ≥3–4 servings/week of choc- =0.002), but results for diabetics were based on a small number
olate had a 20% (95% CI 10% to 30%) lower risk of MI com- of individuals (n=3434) and were not statistically significant.
pared with non-consumers (table 2). The association was The relation between chocolate consumption and MI was not
attenuated, but remained statistically significant after further modified by high education, overweight or history of hyperten-
adjustment for BMI and history of diabetes, hypertension and sion or hypercholesterolaemia ( p for interaction between choc-
hypercholesterolaemia at baseline; the corresponding reduction olate intake and the potential effect modifier in relation to
in MI risk was 13% (95% CI 2% to 23%). The association MI risk >0.25 for all).
between chocolate consumption and MI risk was not modified
by sex ( p for interaction between chocolate consumption and Meta-analysis
sex in relation to MI =0.38). Excluding cases of MI diagnosed Our literature search identified five prospective studies investigat-
during the first 3 years of follow-up did not change the results, ing the association between chocolate consumption and risk of
but widened the CI owing to fewer cases in the analysis IHD (see online supplementary eFigure 1). Together with the
(n=3551) (RR 0.87, 95% CI 0.76 to 1.00, for the highest vs Swedish study, six studies from five different countries (two from
lowest category). Sweden and one each from Germany, the UK, the USA and
In stratified analysis, the association between chocolate con- Australia) with a total of 6851 IHD cases ascertained among
sumption and MI risk appeared to be stronger in individuals 144 823 adults were included in the meta-analysis (table 3). Only
Table 2 RR (95% CI) of MI by chocolate consumption in 67 640 Swedish men and women, 1998−2010
Categories of chocolate consumption, servings
MI cases*/participants (%) 833/10 037 (8.3%) 2142/34 231 (6.3%) 1049/17 487 (6.0%) 393/5885 (6.7%)
Total person-years 116 942 413 046 211 491 69 317
Age-adjusted and sex-adjusted 1.00 0.82 (0.76 to 0.89) 0.77 (0.70 to 0.84) 0.76 (0.67 to 0.85) <0.001
Multivariable model 1† 1.00 0.85 (0.79 to 0.92) 0.82 (0.75 to 0.90) 0.80 (0.70 to 0.90) 0.002
Multivariable model 2‡ 1.00 0.91 (0.84 to 0.99) 0.89 (0.81 to 0.97) 0.87 (0.77 to 0.98) 0.04
*Non-fatal and fatal cases.
†The Cox proportional hazards regression model is stratified by baseline age (in years) and sex and includes education (less than high school, high school, university), family history of
MI before 60 years of age (no, yes), smoking (never; past <20 or ≥20 pack-years; current <20 or ≥20 pack-years), aspirin use (never, 1–6 tablets/week, ≥7 tablets/week), walking/
bicycling (almost never, <20 min/day, 20–40 min/day, 40–60 min/day, >1 h/day), exercise (<1 h/week, 1 h/week, 2–3 h/week, 4–5 h/week, >5 h/week) and intakes of total energy (kcal/
day; continuous), alcohol (g/day; quintiles), processed meat (servings/week; quintiles) and fruits and vegetables (servings/day; quintiles).
‡Adjusted for the same covariates as above and further for body mass index (kg/m2; continuous) and history and diagnosis of diabetes (no, yes), hypertension (no, yes) and
hypercholesterolaemia (no, yes).
MI, myocardial infarction; RR, relative risk.
Mink et al7
Iowa Women’s Health 34 489 CVD-free 16 1329 fatal IHD 0 1.00 (reference) Age, marital status, education, blood pressure, diabetes, BMI,
Study, USA postmenopausal women, cases >0 serving/week 0.98 (0.88 to 1.10) waist-to-hip ratio, physical activity, smoking, oestrogen use and energy
55–69 years intake
Janszky et al8 Stockholm Heart 1169 non-diabetic men and 8 250 recurrent Never 1.00 (reference) Age, sex, education, smoking, obesity, physical inactivity, alcohol
Epidemiology Program, women with MI, 45–70 years non-fatal MI <1 serving/month 0.95 (0.61 to 1.49) consumption, filtered coffee consumption and sweet score
Sweden cases 1 serving/week 1.02 (0.68 to 1.55)
≥2 servings/week 0.86 (0.54 to 1.37)
Buijsse et al9 EPIC-Potsdam, Germany 19 357 CVD-free men and 8.3 166 non-fatal 11.9 g/week 1.00 (reference) Age, sex, employment status, education, smoking, BMI, waist
women not using and fatal MI 13.3 g/week 0.65 (0.40 to 1.05) circumference, prevalence of diabetes, occupational physical activity,
antihypertensive medication, cases 23.1 g/week 1.02 (0.65 to 1.60) sports, cycling and intakes of total energy, alcohol, coffee, tea, red
35–65 years 52.5 g/week 0.73 (0.47 to 1.15) meat, processed meat, dairy, fruits, vegetables and cereal fibre
Lewis et al10 NA, Australia 1216 women, NA 9.5 153 non-fatal <1 serving/week 1.00 (reference) Age, socioeconomic status, BMI and energy intake
and fatal IHD ≥1 serving/week 0.65 (0.46 to 0.94)
cases
Larsson SC, et al. Heart 2016;102:1017–1022. doi:10.1136/heartjnl-2015-309203
Kwok et al11 EPIC-Norfolk, UK 20 952 CVD-free men and 11.9 2434 non-fatal 0 1.00 (reference) Age, sex, smoking, physical activity, BMI, diabetes, systolic blood
women, mean age 59 years and fatal IHD 4.2–24.5 g/week 1.03 (0.91 to 1.15) pressure, low-density lipoprotein cholesterol, high-density lipoprotein
cases 28.7–49.9 g/week 1.03 (0.91 to 1.17) cholesterol and intake of energy and alcohol
50.4–108.5 g/week 0.92 (0.81 to 1.05)
≥109.2 g/week 0.91 (0.80 to 1.04)
Larsson et al, Cohort of Swedish Men 67 640 CVD-free men and 13 4417 non-fatal 0 1.00 (reference) Age, sex, education, family history of MI, smoking status and
2016 (current and Swedish women, 45–83 years and fatal MI 1–3/month 0.91 (0.84 to 0.99) pack-years of smoking, aspirin use, walking/bicycling, exercise, BMI,
study) Mammography Cohort, cases 1–2/week 0.89 (0.81 to 0.97) history of diabetes, hypertension and hypercholesterolaemia, intakes of
Sweden ≥3–4/week 0.87 (0.77 to 0.98) total energy, alcohol, processed meat and fruits and vegetables
BMI, body mass index; CVD, cardiovascular disease; EPIC, European Prospective Investigation into Cancer; IHD, ischaemic heart disease; MI, myocardial infarction; NA, not available; RR relative risk.
Downloaded from http://heart.bmj.com/ on December 20, 2016 - Published by group.bmj.com
Figure 1 Forest plot for the meta-analysis of chocolate consumption (highest vs lowest category) and risk of myocardial infarction or ischaemic
heart disease. Squares indicate study-specific RR (size of the square reflects the study-specific statistical weight); the horizontal lines indicate 95%
CIs; diamond indicates the overall RR with its 95% CI. *These studies have multiple categories for chocolate consumption; the other studies have
only two categories (see table 3). COSM, Cohort of Swedish Men; EPIC, European Prospective Investigation into Cancer; IWHS, Iowa Women’s
Health Study; NA, not available; RR, relative risk; SHEEP, Stockholm Heart Epidemiology Program; SMC, Swedish Mammography Cohort.
two studies defined type of chocolate consumption.9 11 In the MI or IHD. Because of the large number of MI cases, we had
Postdam arm of the European Prospective Investigation into high statistical power to detect an association.
Cancer (EPIC) study, chocolate consumption included chocolate A beneficial effect of chocolate consumption on risk of MI/
bars.9 In the EPIC-Norfolk cohort, chocolate consumption IHD is biologically plausible. Findings from a meta-analysis of
included plain chocolate, chocolate snack bars and cocoa powder several short-term randomised controlled trials showed that
(for cocoa drinks).11 When results from all six studies were com- interventions with chocolate or cocoa significantly improved
bined, the overall RR for the highest versus lowest category of insulin sensitivity and endothelial function as well as reduced
chocolate consumption was 0.90 (95% CI 0.82 to 0.97), with fasting insulin concentration, diastolic blood pressure and mean
little between-study heterogeneity (I2=24.3%) (figure 1). arterial pressure.1 Chocolate or cocoa flavanol intake was also
Excluding the Swedish study did not alter the results appreciably, found to increase high-density lipoprotein cholesterol and to
although the CI was somewhat widened (RR, 0.89; 95% CI 0.79 decrease low-density lipoprotein cholesterol and triglycerides.1
to 1.01). We found no statistically significant publication bias A randomised trial conducted in 90 individuals showed that
( p=0.12). those who were allocated to daily consume a drink high in
The current Swedish study and three other studies8 9 11 cocoa flavanols for 8 weeks had statistically significant improve-
reported results for three or more categories of chocolate con- ments in insulin sensitivity, blood pressure and lipid profile as
sumption and were included in the dose–response meta-analysis. well as reduced lipid peroxidation and plasma glucose and
The overall RR per 50 g/week increment of chocolate consump- insulin concentrations compared with the control group who
tion was 0.95 (95% CI 0.92 to 0.98), without heterogeneity consumed a drink low in flavanols.2 Another recent randomised
among studies (I2=0%) (see online supplementary eFigures 2 controlled trial showed that a high cocoa flavanol intake, com-
and 3). pared with control, increased flow-mediated vasodilation by
1.2%, decreased systolic and diastolic blood pressure by respect-
ively 4.4 and 3.9 mmHg, decreased pulse wave velocity by
DISCUSSION 0.4 m/s and improved blood lipid concentrations.3 Not all trials,
In this large prospective study of Swedish adults, high chocolate however, have found significant effects of cocoa flavanols on
consumption was associated with a significant 13% reduced risk blood pressure.18
of MI. In a complementary meta-analysis, including the Swedish Major strengths of the current study of Swedish adults are the
study and five published prospective studies, we observed that a large sample size, the large number of incident MI cases, the pos-
high consumption of chocolate was associated with a significant sibility to adjust for major potential confounders and the object-
10% lower risk of IHD. ive data on MI diagnoses obtained through linkage with Swedish
Although several studies have examined the association registries. A limitation is that chocolate consumption was assessed
between chocolate consumption and risk of total IHD incidence with only one question in the FFQ and the lack of information
or mortality,7 10 11 only one previous study has assessed the on type of chocolate (eg, plain chocolate or chocolate bars).
relation between chocolate consumption and risk of MI in a Moreover, we could not distinguish between dark chocolate and
population without previous MI.9 That study included 166 MI milk chocolate. An inverse association between chocolate con-
cases and found a statistically non-significant 27% lower risk of sumption and MI is assumed to be stronger for dark chocolate,
MI when comparing the highest with the lowest quartile of which has higher cocoa content than milk chocolate. Swedish
chocolate consumption.9 The present study of Swedish adults is milk chocolate normally contains at least 30% cocoa solids.8
the largest study to date on chocolate consumption and risk of Although we had no information on the proportion of dark
Larsson SC, et al. Heart 2016;102:1017–1022. doi:10.1136/heartjnl-2015-309203 1021
Downloaded from http://heart.bmj.com/ on December 20, 2016 - Published by group.bmj.com
versus milk chocolate intake in the whole study population, in a Funding This work was supported by the Swedish Research Council and by a
subsample of participants who had completed an extensive FFQ Young Scholars Award grant from the Strategic Research Area in Epidemiology at
Karolinska Institutet.
in 2010, 54% of participants reported that they consumed more
dark chocolate than milk chocolate (unpublished data). Another Competing interests None declared.
limitation is that chocolate intake was self-reported and measured Patient consent Obtained.
at baseline only, which may have resulted in some misclassifica- Ethics approval Regional Ethical Review Board at Karolinska Institutet, Stockholm,
tion of long-term chocolate intake. In a reproducibility study Sweden.
conducted in subsample of women from the study cohort, the Provenance and peer review Not commissioned; externally peer reviewed.
reliability of chocolate consumption assessed by two FFQs admi-
nistered 1 year apart was found to be relatively high (correlation REFERENCES
coefficient for the unadjusted data was 0.65; unpublished data). 1 Hooper L, Kay C, Abdelhamid A, et al. Effects of chocolate, cocoa, and flavan-3-ols
Because the study population comprised middle-aged and older on cardiovascular health: a systematic review and meta-analysis of randomized
trials. Am J Clin Nutr 2012;95:740–51.
Caucasian men and women, our findings may not be generalis- 2 Mastroiacovo D, Kwik-Uribe C, Grassi D, et al. Cocoa flavanol consumption
able to younger individuals and other ethnic groups. Finally, as in improves cognitive function, blood pressure control, and metabolic profile in elderly
any observational study, we cannot rule out the possibility of subjects: the Cocoa, Cognition, and Aging (CoCoA) Study—a randomized
residual confounding. controlled trial. Am J Clin Nutr 2015;101:538–48.
3 Sansone R, Rodriguez-Mateos A, Heuel J, et al. Cocoa flavanol intake improves
endothelial function and Framingham Risk Score in healthy men and women: a
CONCLUSIONS randomised, controlled, double-masked trial: the Flaviola Health Study. Br J Nutr
In summary, results from this Swedish prospective study and a 2015;114:1246–55.
complementary meta-analysis indicate that chocolate consump- 4 Larsson SC, Virtamo J, Wolk A. Chocolate consumption and risk of stroke: a
tion is inversely associated with risk of MI and IHD. Although prospective cohort of men and meta-analysis. Neurology 2012;79:1223–9.
5 Zhang Z, Xu G, Liu X. Chocolate intake reduces risk of cardiovascular disease:
chocolate consumption may lower the IHD risk, chocolate evidence from 10 observational studies. Int J Cardiol 2013;168:5448–50.
should be consumed in moderation because it is high in sugar, 6 Buitrago-Lopez A, Sanderson J, Johnson L, et al. Chocolate consumption and
calories and saturated fat. cardiometabolic disorders: systematic review and meta-analysis. BMJ 2011;343:d4488.
7 Mink PJ, Scrafford CG, Barraj LM, et al. Flavonoid intake and cardiovascular disease
mortality: a prospective study in postmenopausal women. Am J Clin Nutr
2007;85:895–909.
8 Janszky I, Mukamal KJ, Ljung R, et al. Chocolate consumption and mortality
Key messages following a first acute myocardial infarction: the Stockholm Heart Epidemiology
Program. J Intern Med 2009;266:248–57.
9 Buijsse B, Weikert C, Drogan D, et al. Chocolate consumption in relation to blood
What is already known on this subject? pressure and risk of cardiovascular disease in German adults. Eur Heart J
Available evidence indicates beneficial effects of chocolate and 2010;31:1616–23.
cocoa flavanols on cardiovascular health. Furthermore, 10 Lewis JR, Prince RL, Zhu K, et al. Habitual chocolate intake and vascular disease: a
prospective study of clinical outcomes in older women. Arch Intern Med
prospective studies have reported an inverse association 2010;170:1857–8.
between chocolate consumption and stroke incidence and 11 Kwok CS, Boekholdt SM, Lentjes MA, et al. Habitual chocolate consumption and
cardiovascular disease mortality. risk of cardiovascular disease among healthy men and women. Heart
2015;101:1279–87.
What might this study add? 12 Messerer M, Johansson SE, Wolk A. The validity of questionnaire-based
In this prospective study of 67 640 Swedish adults, chocolate micronutrient intake estimates is increased by including dietary supplement use in
consumption was significantly inversely associated with risk of Swedish men. J Nutr 2004;134:1800–5.
13 Greenland S, Longnecker MP. Methods for trend estimation from summarized
myocardial infarction. In a complementary meta-analysis, high dose-response data, with applications to meta-analysis. Am J Epidemiol
consumption of chocolate was associated with a 10% lower risk 1992;135:1301–9.
of ischaemic heart disease. 14 Orsini N, Bellocco R, Greenland S. Generalized least squares for trend estimation of
summarized dose-response data. Stata J 2006;6:40–57.
How might this impact on clinical practice? 15 DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials
Chocolate consumption in moderation might lower the risk of 1986;7:177–88.
16 Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat Med
ischaemic heart disease. 2002;21:1539–58.
17 Egger M, Davey Smith G, Schneider M, et al. Bias in meta-analysis detected by a
Contributors SCL analysed the data, conducted the literature review and wrote the simple, graphical test. BMJ 1997;315:629–34.
manuscript. AW participated in the data collection. SCL, AÅ, BG and AW interpreted 18 Ottaviani JI, Balz M, Kimball J, et al. Safety and efficacy of cocoa flavanol intake in
the data and critically reviewed the paper. All authors read and approved the final healthy adults: a randomized, controlled, double-masked trial. Am J Clin Nutr
manuscript. 2015;102:1425–35.
These include:
References This article cites 18 articles, 10 of which you can access for free at:
http://heart.bmj.com/content/102/13/1017#BIBL
Email alerting Receive free email alerts when new articles cite this article. Sign up in the
service box at the top right corner of the online article.
Notes