The strongest human evidence on cocoa and health. Nearly all the randomised trials tested flavanol extracts or specially made high-flavanol products, not ordinary chocolate, and every summary says which. This is a reading list with its limits shown; how evidence is graded on the health pages is a separate and stricter exercise.
26 studies1984 to 2026Each checked against its published abstract
Nothing here is medical advice. The one health claim authorised for cocoa in the European Union reads “Cocoa flavanols help maintain the elasticity of blood vessels, which contributes to normal blood flow”, requires 200 mg of cocoa flavanols a day, and was granted in 2013 on an application by a manufacturer whose own study the assessment depended on (Commission Regulation (EU) No 851/2013). For the health pages themselves, see health.
What these studies add up to
The largest trial did not find what the smaller ones implied. Over 21,000 people took a cocoa extract with 500 mg of flavanols daily for years: the main cardiovascular endpoint was not significantly reduced, and there was no effect on cognition, type 2 diabetes or macular degeneration.
Short trials do show small changes in things measured in a clinic: blood pressure down by under 2 mmHg, artery dilation up by about one percentage point, modest shifts in insulin and blood fats. Reviewers graded this evidence low to moderate and flagged risk of bias.
Studies that follow chocolate eaters over years find them with somewhat less heart disease and diabetes. These are observations, not experiments, and an umbrella review graded the evidence weak.
On weight the three kinds of study give three answers: surveys find chocolate eaters leaner, a cohort found more weight gain with more chocolate, and trials found no change.
Chocolate as a migraine trigger did not survive a blinded test against carob, though one patient in seven believes it.
Theobromine, chocolate's main stimulant, did little to mood or alertness at the doses found in food.
The trials tested flavanols at doses that heavily alkalised cocoa and most ordinary chocolate do not supply.
In 79 cocoa and chocolate foods, theobromine was ten to twenty times as abundant as caffeine
How much caffeine and theobromine do common household foods containing cocoa or chocolate contain?
How
Analytical survey, no human participants: 79 commonly used household foodstuffs containing cocoa or chocolate, extracted with boiling water and measured by reversed-phase high-performance liquid chromatography.
Finding
Mean caffeine levels in cocoa cereals, chocolate bakery products, chocolate toppings, cocoa beverages, chocolate ice creams and chocolate milks were 0.071, 0.152, 0.138, 0.208, 0.032 and 0.011 mg/g. Mean theobromine levels were 0.695, 1.47, 1.95, 2.66, 0.621 and 0.226 mg/g. Chocolate puddings held 74.8 mg theobromine and 6.5 mg caffeine per serving.
Does not establish: These figures are for chocolate-flavoured foods, most of them dilute, and do not give the caffeine or theobromine content of a chocolate bar. They show the ratio between the two compounds, not a dose that would affect a person.
Limits: US products sold in the early 1980s; the abstract reports composite foods, not plain dark or milk chocolate bars.
Does chocolate provoke attacks in migraine patients who believe it does?
How
Intervention was chocolate against a closely matching placebo. Double-blind parallel group study in patients with migraine who believed that chocolate could provoke their attacks: 12 received chocolate and 8 placebo. The amount of chocolate is not given in the abstract.
Finding
Chocolate ingestion was followed by a typical migraine episode in 5 out of 12 patients, while none of the 8 patients challenged with placebo had an attack (p = 0.051). The median time to the onset of the attack was 22 h.
Does not establish: This does not show that chocolate is a migraine trigger in general. It is a borderline result in 20 selected patients, and the larger blinded study by Marcus and colleagues did not reproduce it.
Limits: Twenty patients in total; the result does not reach the conventional significance threshold; participants were selected for believing they were sensitive.
Is chocolate more likely than a carob placebo to provoke headache in women with chronic headache?
How
Intervention was chocolate with carob as placebo. Double-blind provocative study in sixty-three women with chronic headache (50% migraine, 37.5% tension-type, 12.5% combined). After 2 weeks on a diet restricting vasoactive amine-rich foods, each subject received two samples of chocolate and two of carob in random order and kept diaries. The amount of chocolate is not given in the abstract.
Finding
Chocolate was not more likely to provoke headache than carob in any of the headache diagnostic groups (chi2(2)=0.36, p=0.83). The result was independent of the subjects' own beliefs about chocolate as a trigger (chi2(1)=0.73, p=0.39).
Does not establish: The study does not show that chocolate never triggers a headache in anyone. It shows that in a blinded test of typical headache sufferers, including those who believed chocolate was a trigger, chocolate did no worse than placebo.
Limits: Women only; a single formulation and dose; a sample of 63 cannot exclude a trigger effect in a small minority.
Is chocolate consumption associated with the risk of developing cardiometabolic disorders?
How
Exposure was self-reported chocolate consumption, not an intervention. Systematic review and meta-analysis: from 4576 references seven studies met the inclusion criteria, including 114,009 participants; six were cohort studies and one cross sectional. Comparison was highest against lowest level of chocolate consumption.
Finding
The highest levels of chocolate consumption were associated with a 37% reduction in cardiovascular disease (relative risk 0.63, 95% confidence interval 0.44 to 0.90) and a 29% reduction in stroke compared with the lowest levels. None of the studies was a randomised trial, and large variation was observed between the seven studies in how chocolate consumption was measured.
Does not establish: This is an association, not a demonstrated effect. The paper does not show that eating more chocolate lowers anyone's risk, and its own conclusion asks for experimental studies to confirm a benefit.
Limits: Observational only; chocolate intake was measured differently in each study; people who eat more chocolate may differ in health, income and diet in ways adjustment does not remove.
What are the effects of chocolate, cocoa and flavan-3-ols on major cardiovascular risk factors?
How
Interventions were chocolate, cocoa or flavan-3-ols, pooled together. Systematic review and random-effects meta-analysis of 42 acute or short-term chronic (18 wk or less) randomised controlled trials comprising 1297 participants.
Finding
HOMA-IR improved (-0.67; 95% CI -0.98 to -0.36). Flow-mediated dilatation improved after chronic (1.34%; 95% CI 1.00% to 1.68%) and acute (3.19%; 95% CI 2.04% to 4.33%) intakes. Diastolic blood pressure fell by 1.60 mm Hg (95% CI -2.77 to -0.43). Effects on LDL (-0.07 mmol/L) and HDL (0.03 mmol/L) cholesterol were marginally significant. GRADE suggested low- to moderate-quality evidence.
Does not establish: The review pooled chocolate, cocoa drinks and isolated flavan-3-ols, so its averages do not describe chocolate bars. It measured risk markers, not heart attacks, and its own conclusion is that larger, longer and independently funded trials are needed.
Limits: No trial exceeded 18 weeks. Evidence strength was lowered for unclear allocation concealment, dropouts, missing outcome data and heterogeneity. The authors ask for independently funded trials.
Does the cross-sectional association between chocolate intake and lower body weight hold in a prospective analysis?
How
Exposure was self-reported frequency of eating a 1-oz (about 28 g) serving of chocolate, not an intervention. Prospective analysis of the Atherosclerosis Risk in Communities cohort: 15,732 and 12,830 participants at the first and second visit, six years apart; weight and height were measured.
Finding
Compared with participants who ate a chocolate serving less often than monthly, those who ate it 1-4 times a month and at least weekly had an increase in body mass index of 0.26 (95% CI 0.08 to 0.44) and 0.39 (0.23 to 0.55) kg/m2 over the six-year period. In cross-sectional analyses chocolate frequency was inversely associated with body weight, and this was attenuated after excluding participants with pre-existing obesity-related illness.
Does not establish: The study does not prove chocolate causes weight gain, but it shows why cross-sectional findings that chocolate eaters are slimmer cannot be read as chocolate keeping people slim: people who were already ill reported eating less chocolate.
Limits: Observational; chocolate type not distinguished; intake self-reported; missing data were imputed.
What are the subjective and cardiovascular effects of a range of oral theobromine doses in healthy people?
How
Intervention was isolated theobromine, not chocolate: oral doses of 250, 500 and 1,000 mg, with caffeine 200 mg as active control and placebo, in a within-subjects study of 80 healthy participants.
Finding
Theobromine showed limited subjective effects at 250 mg and negative mood effects at higher doses, and dose-dependently increased heart rate. Caffeine had the expected effects on alertness and cardiovascular parameters. The abstract gives no effect sizes.
Does not establish: The study does not show that theobromine explains why people like chocolate; the authors say only that it may contribute at normal intake ranges. Smit and Blackburn (2005) put the theobromine in a 50-g bar of chocolate at 250 mg, so the doses that produced negative effects exceed a normal portion.
Limits: Single doses of the pure compound in healthy volunteers; no magnitudes in the abstract.
Does theobromine at doses commonly found in foods affect mood and vigilance in the way caffeine does?
How
Intervention was theobromine in a beverage, not chocolate: twenty-four men completed 6 double-blind trials with 100, 200 and 400 mg theobromine in a cocoa-based beverage and matched controls (0 mg theobromine, 400 mg theobromine and 100 mg caffeine) in a non-cocoa beverage, followed by mood questionnaires and a 2-hour visual vigilance task.
Finding
Theobromine did not influence mood or vigilance at these doses. Caffeine at 100 mg significantly decreased lethargy/fatigue and increased vigor (P = 0.006 and 0.011, respectively). Salivary theobromine rose in a dose-dependent manner, confirming absorption.
Does not establish: The study does not show that chocolate has no effect on alertness, since chocolate also contains some caffeine and sugar. It shows that theobromine alone, at food-level doses, is not a meaningful stimulant.
Limits: Twenty-four young men, single acute doses, one vigilance task.
Does chocolate exacerbate acne in men aged 18 to 35 with a history of acne vulgaris?
How
Intervention was cocoa powder in capsules, not chocolate: fourteen men aged 18 to 35 swallowed capsules filled with unsweetened 100-percent cocoa, hydrolysed gelatin powder, or a combination, at baseline. Double-blind, placebo-controlled, randomised; lesions assessed at baseline, Day 4 and Day 7; 13 completed. The cocoa dose is not given in the abstract.
Finding
A statistically significant increase in the mean number of total acneiform lesions was detected on Day 4 (p=0.006) and Day 7 (p=0.043) compared to baseline. The correlation between amount consumed and number of lesions was r=0.250 to Day 4 and r=0.314 to Day 7.
Does not establish: With 13 participants and weak correlations, this does not establish that chocolate causes acne. It tested pure cocoa without sugar or milk, so it cannot say which part of a chocolate bar, if any, matters.
Limits: Thirteen completers; the reported comparison is against baseline and a dose correlation, not a between-group difference; single dose; one week of follow-up; men only.
What is the effect of cocoa flavanol intake on cardiometabolic biomarkers in randomised trials?
How
Interventions were cocoa flavanols in various forms, 166 to 2110 mg/d, for 2 to 52 wk. Systematic review and random-effects meta-analysis of 19 randomised controlled trials comprising 1131 participants; the number of studies for a specific biomarker varied.
Finding
Weighted mean differences between treatment and placebo were -0.10 mmol/L (95% CI -0.16 to -0.04) for triglycerides, 0.06 mmol/L (95% CI 0.02 to 0.09) for HDL cholesterol, -2.33 μIU/mL (95% CI -3.47 to -1.19) for fasting insulin and -0.93 (95% CI -1.31 to -0.55) for HOMA-IR. No significant associations were found for other biomarkers.
Does not establish: The meta-analysis did not show that cocoa prevents diabetes or cardiovascular events; it called for large long-term trials, and the later COSMOS trial (Li and colleagues) found that cocoa extract did not reduce diabetes.
Limits: Trials were small and mostly short. Results are for biomarkers, and the authors state that large long-term trials were still lacking at the time.
Twenty-five men eating very dark chocolate daily had worse acne, with no control group
Vongraviopap and Asawanonda, 2016. International Journal of Dermatology. doi:10.1111/ijd.13188
Question
Does 99% dark chocolate, eaten in regular daily amounts, worsen acne in acne-prone men?
How
Intervention was chocolate: 25 g of 99% dark chocolate daily for 4 weeks in twenty-five acne-prone male subjects, assessed weekly by Leeds revised acne scores and lesion counts. Before-and-after comparison; the abstract describes no control group or blinding.
Finding
Statistically significant changes in acne scores and in numbers of comedones and inflammatory papules were detected as early as 2 weeks into the study and remained significant at 4 weeks compared to baseline. The abstract gives no lesion counts.
Does not establish: An uncontrolled before-and-after study cannot separate an effect of chocolate from natural variation or from participants and assessors knowing what was eaten. It says nothing about milk chocolate, women, or people without acne-prone skin.
Limits: No control group, no blinding, men only, small sample, and acne fluctuates naturally over four weeks.
What is the effect on blood pressure of chocolate or cocoa products compared with low-flavanol products or placebo, taken for two weeks or longer?
How
Interventions were chocolate or cocoa products, mixed across trials: 30 to 1218 mg of flavanols (mean 670 mg) in 1.4 to 105 grams of cocoa products per day. Cochrane systematic review and random-effects meta-analysis of randomised controlled trials: thirty-five trials with 40 treatment comparisons, 1804 mainly healthy participants, trial duration two to 18 weeks.
Finding
Flavanol-rich cocoa products reduced systolic blood pressure by a mean of 1.76 mmHg (95% CI -3.09 to -0.43, P = 0.009) and diastolic by 1.76 mmHg (95% CI -2.57 to -0.94) compared with control. A sensitivity analysis excluding trials with authors employed by the sponsoring industry (33 trials, 1482 participants) revealed a small reduction in effect size, indicating some reporting bias. The quality of evidence was downgraded from high to moderate.
Does not establish: The review does not show that eating chocolate treats high blood pressure or prevents cardiovascular events. The mean flavanol dose in the trials (670 mg) is more than three times the 200 mg daily intake on which the EU health claim is conditioned, and the trials measured blood pressure for weeks, not disease over years.
Limits: Trials lasted at most 18 weeks. Heterogeneity could not be explained. A trend towards larger effects in unblinded trials suggests participant expectation may contribute.
What is the dose-response relationship between cocoa flavanol consumption and endothelial function measured by flow-mediated dilation?
How
Interventions were cocoa flavanols in various forms; intervention groups received 80 to 1248 mg (mean 704 mg) more flavanols than control groups. Systematic review and meta-analysis of human intervention studies found in Scopus: fifteen published articles with 18 intervention arms. The abstract gives no participant total.
Finding
Flow-mediated dilation improved by 1.17% (95% CI 0.76% to 1.57%), with a non-linear, inverted U-shaped association. The optimal effect was observed with 710 mg total flavanols, 95 mg (-)-epicatechin or 25 mg (+)-catechin. There was substantial unexplained variation and significant risk-of-bias concerns with a large majority of the studies.
Does not establish: A change in a laboratory measure of artery dilation is not a reduction in heart disease. The effect peaked at a flavanol dose more than three times the 200 mg on which the EU health claim is conditioned, so the result cannot be read as a property of chocolate bars as sold.
Limits: Only one database was searched; most included trials had risk-of-bias concerns; flow-mediated dilation is a surrogate marker.
What is the dose-response association between chocolate consumption and incident cardiovascular disease?
How
Exposure was self-reported chocolate consumption, not an intervention. Dose-response meta-analysis of prospective studies: fourteen publications (23 studies including 405 304 participants and 35 093 cases of CVD), searched to 6 June 2018.
Finding
The summary relative risk per 20 g/week increase in chocolate consumption was 0.982 (95% CI 0.972 to 0.992) for CVD and 0.956 (0.932 to 0.980) for total stroke; for heart failure it was 0.995 (0.981 to 1.010). The dose-response was non-linear, with the lowest risk at 45 g/week (RR 0.890; 95% CI 0.849 to 0.932). The authors conclude an association with reduced risk at under 100 g/week.
Does not establish: The paper does not show that 45 g a week is a protective dose. A dip in risk at low intake in cohort data is what confounding by general health and lifestyle would also produce, and the authors themselves warn that higher intakes may bring harm from sugar.
Limits: Observational; chocolate type is not distinguished in the abstract; intake comes from food frequency questionnaires; residual confounding cannot be excluded.
How strong is the evidence linking chocolate consumption with health outcomes across published systematic reviews?
How
Umbrella review of systematic reviews of observational and intervention studies of chocolate consumption: from 240 articles, 10 systematic reviews (8 with a meta-analysis) covering 84 studies (36 prospective observational, 48 interventional) and nineteen outcomes. Observational studies included 1,061,637 participants.
Finding
In observational studies chocolate consumption was associated with reduced risk of cardiovascular death, acute myocardial infarction, stroke and diabetes, on weak evidence of credibility. In intervention studies it was associated with flow-mediated dilatation at 90-150 min and at 2-18 weeks and with insulin resistance markers, with GRADE evidence low or very low. Two systematic reviews reported no association with better depressive mood or cognitive function.
Does not establish: The review does not conclude that chocolate is good for health. Its own summary is that there is weak evidence that chocolate consumption may be associated with favourable outcomes.
Limits: An umbrella review inherits the limits of the reviews it summarises; the abstract does not report effect sizes.
Does cocoa or dark chocolate supplementation affect body weight, body mass index and waist circumference?
How
Interventions were cocoa or dark chocolate, pooled. Systematic review, random-effects meta-analysis and dose-response analysis of 35 randomised clinical trials, searched up to December 2017. The abstract gives no participant total.
Finding
No significant effect on body weight (-0.108 kg, 95% CI -0.262 to 0.046, P = 0.168), BMI (-0.014 kg/m2, 95% CI -0.105 to 0.077) or waist circumference (0.025 cm, 95% CI -0.083 to 0.129, P = 0.640). A subgroup analysis found weight and BMI reduced with 30 g or more of chocolate per day in trials lasting between 4 and 8 weeks.
Does not establish: The review does not show that dark chocolate helps weight loss. The overall result is null and the subgroup signal is small and exploratory. Short trials designed around other outcomes do not describe what happens when people add chocolate to their usual diet for years.
Limits: Trials were short and were designed to test other outcomes; the subgroup finding is one of several subgroup comparisons.
Of 156 dogs brought in after eating chocolate, 44 showed poisoning and 43 of those survived
Weingart and colleagues, 2021. The Journal of Small Animal Practice. doi:10.1111/jsap.13329
Question
What are the clinical features and outcome of dogs after chocolate ingestion?
How
Veterinary case series, not a human study: retrospective evaluation of 156 dogs after chocolate ingestion. Methylxanthine dose (theobromine, caffeine) was calculated from the type and amount of chocolate; the abstract reports no dose figures.
Finding
One hundred and twelve dogs had no clinical signs; forty-four had signs of chocolate intoxication, and twenty-eight of these 44 had eaten dark and bitter chocolate. Presenting signs included agitation (33), tremor (22) and vomiting (21); sinus tachycardia was found in 28. After decontamination and symptomatic treatment, 43 of the 44 dogs survived, a mortality rate of less than 3%.
Does not establish: The abstract gives no toxic or lethal dose, so it cannot be used to say how much chocolate is dangerous for a dog of a given weight. The good outcome followed prompt veterinary treatment and does not mean chocolate ingestion can be left untreated.
Limits: One referral clinic; retrospective; doses were estimated from owners' accounts; dogs that were never brought in are not counted.
The largest trial: cocoa extract did not significantly reduce cardiovascular events
Sesso and colleagues, 2022. The American Journal of Clinical Nutrition. doi:10.1093/ajcn/nqac055
Question
Does a daily cocoa extract supplement reduce total cardiovascular disease events in older adults?
How
Intervention was a cocoa flavanol extract in supplement form, not chocolate or cocoa powder. Randomised, double-blind, placebo-controlled factorial trial (COSMOS) in 21,442 US adults (12,666 women aged 65 and over, 8776 men aged 60 and over) free of major cardiovascular disease. Dose 500 mg flavanols per day including 80 mg (-)-epicatechin; median follow-up 3.6 years.
Finding
Confirmed total cardiovascular events occurred in 410 participants on cocoa extract and 456 on placebo (HR 0.90; 95% CI 0.78 to 1.02; P = 0.11), which is not statistically significant. Cardiovascular death, a secondary endpoint, had HR 0.73 (95% CI 0.54 to 0.98). All-cause mortality had HR 0.89 (95% CI 0.77 to 1.03). A per-protocol analysis gave HR 0.85 (95% CI 0.72 to 0.99) for total cardiovascular events.
Does not establish: The trial tested a concentrated extract capsule delivering a fixed flavanol dose. It says nothing direct about eating chocolate, which was not tested and which comes with sugar and fat, and it does not show that chocolate prevents heart attacks or strokes.
Limits: The primary outcome was null; the cardiovascular death result is a secondary endpoint and the per-protocol result is not the randomised comparison. Participants were older US adults followed for a median of under four years.
Do cocoa-derived foods affect depressive and anxiety symptoms and positive and negative affect?
How
Interventions were cocoa-rich products of several kinds (chocolate and cocoa drinks pooled). Systematic review and random-effects meta-analysis: 761 records screened, nine studies selected. Two trials lasted more than 1 week, two lasted 3 days and five tested a single administration. The abstract gives no participant total.
Finding
Pooled effects favoured cocoa-rich products for depressive symptoms (Hedge's g = -0.42, 95% CI -0.67 to -0.17) and anxiety symptoms (Hedge's g = -0.49, 95% CI -0.78 to -0.19), positive affect (g = 0.41, 95% CI 0.06 to 0.77) and negative affect (g = -0.47, 95% CI -0.91 to -0.03). The authors state the results cannot be generalised to long-term intake.
Does not establish: The review does not show that chocolate treats or prevents depression. The effects are short-term changes in mood ratings, and an umbrella review of earlier evidence found no association between chocolate and better depressive mood.
Limits: Few studies, few participants, mostly single-dose or three-day designs; blinding a chocolate intervention is difficult.
Does daily cocoa extract, or a daily multivitamin, improve cognition in older women and men?
How
Intervention was a cocoa extract supplement containing 500 mg per day of flavanols, not chocolate. COSMOS-Mind, a randomised two-by-two factorial 3-year trial with cognition assessed by telephone at baseline and annually; 2262 participants enrolled, mean age 73 years, 60% women.
Finding
Cocoa extract had no effect on global cognition (mean z-score 0.03, 95% CI -0.02 to 0.08; P = .28). The multivitamin, relative to placebo, produced a statistically significant benefit on global cognition (mean z 0.07, 95% CI 0.02 to 0.12; P = .007).
Does not establish: The study gives no support to the idea that cocoa flavanols, let alone chocolate, protect memory or thinking in older people. The trial tested an extract and did not test chocolate.
Limits: Cognition was measured by telephone tests; 89% of participants were non-Hispanic White, and the authors call for confirmation in a more diverse cohort.
Does cocoa extract supplementation reduce the incidence of type 2 diabetes?
How
Intervention was a cocoa extract supplement (500 mg per day cocoa flavanols, including 80 mg (-)-epicatechin) against placebo, not chocolate. Analysis within the COSMOS randomised trial of 18,381 participants without diabetes at enrolment; intention-to-treat; outcome was self-reported incident type 2 diabetes; median follow-up 3.5 years.
Finding
801 incident cases were reported. Compared with placebo, cocoa extract did not reduce type 2 diabetes (adjusted hazard ratio 1.04, 95% CI 0.91 to 1.20, P = 0.58). The effect was not significantly modified by sex, race, BMI, smoking, physical activity, dietary quality or baseline flavanol status.
Does not establish: Observational studies link chocolate eating with lower diabetes risk, but this randomised test of the flavanols found no reduction. It does not establish that chocolate itself has no effect, because chocolate was not tested, but it removes the main experimental support for a protective mechanism.
Limits: Diabetes was self-reported, participants were older adults, and follow-up was a median of 3.5 years.
A flavanol memory trial missed its main endpoint; a benefit showed only with poorer diets
Brickman and colleagues, 2023. Proceedings of the National Academy of Sciences of the United States of America. doi:10.1073/pnas.2216932120
Question
Does a cocoa flavanol intervention improve hippocampal-dependent memory in older adults, and does the effect depend on habitual diet?
How
Intervention was cocoa extract providing 500 mg of cocoa flavanols per day, against placebo, not chocolate. COSMOS-Web: 3,562 older adults randomly assigned to a 3-y intervention; diet quality assessed in all participants and a urine-based flavanol biomarker in a subset (n = 1,361).
Finding
The prespecified primary end point, an intervention-related improvement in memory in all participants after 1 y, was not statistically significant. The intervention restored memory among participants in the lower tertiles of habitual diet quality or habitual flavanol consumption. The abstract gives no effect sizes.
Does not establish: The headline of the paper speaks of restoring memory, but the trial as a whole did not show a memory benefit. It does not show that people with ordinary diets gain anything, and it tested an extract, not chocolate.
Limits: The positive result comes from subgroups after a null primary endpoint; the abstract reports no magnitudes; the biomarker was measured only in a subset.
What is the effect of cocoa consumption on anthropometric measures, lipids, glycaemic profile and blood pressure in adults?
How
Interventions were cocoa extract or dark chocolate of 70% cocoa or more, for 4 weeks or longer. Systematic review and meta-analysis of randomised clinical trials: thirty-one studies, 1986 participants. Doses are not given in the abstract.
Finding
No effects on body weight, BMI, waist circumference, triglycerides, HDL-c or HbA1c. Reductions in total cholesterol (-8.35 mg/dL, 95% CI -14.01 to -2.69), LDL-c (-9.47 mg/dL, 95% CI -13.75 to -5.20), fasting blood glucose (-4.91 mg/dL, 95% CI -8.29 to -1.52), systolic blood pressure (-2.52 mmHg, 95% CI -4.17 to -0.88) and diastolic blood pressure (-1.58 mmHg, 95% CI -2.54 to -0.62).
Does not establish: The abstract's conclusion speaks of protective effects with clinical impact, but the review measured risk markers over weeks and reports no disease outcomes. Because extract and dark chocolate are pooled, the figures cannot be assigned to chocolate.
Limits: The abstract does not report heterogeneity, risk of bias or how extract trials and chocolate trials differed, and it pools both.
Are dark, milk and total chocolate consumption associated with the risk of type 2 diabetes?
How
Exposure was self-reported chocolate consumption by food frequency questionnaire, not an intervention. Prospective cohort studies (Nurses' Health Study, Nurses' Health Study II, Health Professionals Follow-Up Study): 192 208 participants for total chocolate; 111 654 for chocolate subtypes, assessed from 2006 or 2007.
Finding
Participants consuming 5 or more servings/week of any chocolate had a 10% (95% CI 2% to 17%) lower rate of type 2 diabetes than those who never or rarely consumed chocolate. For dark chocolate at 5 or more servings/week the figure was 21% (5% to 34%). No significant associations were found for milk chocolate. Intake of milk, but not dark, chocolate was positively associated with weight gain.
Does not establish: The study does not show that switching to dark chocolate prevents diabetes. In the COSMOS randomised trial a cocoa extract did not reduce diabetes, and the authors of this paper ask for randomised trials to replicate their finding.
Limits: Observational; participants were health professionals; diabetes was self-reported and confirmed by supplementary questionnaire; dark chocolate eaters may differ from others in ways not captured by adjustment.
Does daily cocoa extract supplementation prevent the development or progression of age-related macular degeneration?
How
Intervention was a cocoa extract supplement (500 mg per day cocoa flavanols, including 80 mg (-)-epicatechin) against placebo, not chocolate. Prespecified ancillary study of the COSMOS randomised trial among 21 442 US adults; median 3.6 years of treatment and follow-up; outcome was self-reported AMD confirmed by medical record review.
Finding
There were 159 cases (1.5%) in the cocoa extract group and 185 cases (1.7%) in the placebo group (HR 0.87; 95% CI 0.71 to 1.08; P = .21). Separate models indicated HR 0.77 (95% CI 0.59 to 1.01) during the first 2 years and HR 1.06 (95% CI 0.76 to 1.50) beyond 2 years.
Does not establish: The trial does not show that cocoa or chocolate protects eyesight. The early-period signal is described by the authors as something that could not be ruled out, not as a demonstrated benefit.
Limits: Only 344 participants had a confirmed AMD event, so the trial had limited power for this outcome; the time-split analysis followed evidence of non-proportional hazards and its early-period interval still includes no effect.
How often do people with migraine report chocolate, caffeine and cheese as triggers?
How
No intervention: systematic review and random-effects meta-analysis of proportions from observational studies of patient-reported triggers. Eight studies included, of which seven contributed to the quantitative analysis; searched to November 11, 2024. The abstract gives no participant total.
Finding
The pooled prevalence of patient-reported migraine triggers was 13.9% for chocolate (95% CI 7.0% to 21.0%), 8.0% for cheese (95% CI 3.0% to 12.0%) and 12.0% for caffeine (95% CI 6.0% to 19.0%), with heterogeneity above 95%.
Does not establish: A patient reporting chocolate as a trigger is not evidence that chocolate caused the attack; the figure measures what patients believe, and the blinded provocation studies test that belief directly. The author concludes that routine avoidance cannot be universally recommended.
Limits: Self-reported, observational, few studies, and very high heterogeneity between them.
We would like to set two Google Analytics cookies to count page views, and to count them in our own cookie-free measurement as well. They do not identify you and nothing is shared with advertisers. Decline and the site works exactly the same — see cookies for what each one does.