Reviewed: Hooper L, Kay C, Abdelhamid A, et al. Effects of chocolate, cocoa, and flavan-3-ols on cardiovascular health: a systematic review and meta-analysis of randomized trials. Am J Clin Nutr. 2012 Mar;95(3):740-751.

Observational and prospective data (i.e., information from long-term, non-controlled studies) have shown evidence that consumption of cocoa (Theobroma cacao, Sterculiaceae) and various types of chocolate preparations are associated with a decrease in cardiovascular disease (CVD) risk and mortality, as well as reduced blood pressure (BP) and cholesterol. Short-term, randomized, controlled trials (RCTs) have brought to light the mechanisms involved in producing these effects. Several systematic reviews and meta-analyses have been performed for different outcomes, but many did not remove lower quality studies or are now out of date. The authors — of the Norwich Medical School, University of East Anglia in Norwich, United Kingdom — have therefore endeavored to perform a systematic analysis of RCTs to assess the effects of cocoa, chocolate, or cocoa flavan-3-ols on classic cardiovascular biomarkers.

The biomarkers included were modifiable Framingham risk measures (as determined in the Framingham study, a large-scale epidemiological study of cardiovascular parameters that commenced in 1948), including systolic and diastolic BP and total, low-density lipoprotein (LDL), and high-density lipoprotein (HDL) cholesterol, as well as independent predictors of CVD risk such as fasting glucose, insulin, triglycerides, hemoglobin A1c (HbA1c), C-reactive protein (CRP), and flow-mediated dilation (FMD). Effects on body mass were also studied. The review was conducted in accordance with Cochrane methodology and presented in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.

Controlled parallel or crossover trials were retrieved from MEDLINE and EMBASE (both on Ovid, www.ovid.com), and the Cochrane Library (CENTRAL, www.thecochranelibrary.com), up to May 2011, using complex, structured searches. For each study found, the following data were gathered: quality characteristics, funding and blinding (to assess potential bias), and the similarity of fat intake between the treatment and control arms. Statistical analyses included the following: tests to assess the variability among studies, sensitivity analyses to assess the robustness of results, and funnel plots (statistical tools to determine if a large or small study comports with averages) to assess the evidence of small study or publication bias. For all outcomes for which there were more than 10 studies, assessments were done for effects of dose, study duration, sex, intervention type, and baseline CVD risk.

Of 1,637 potential studies initially identified, 42 trials (total n=1,297 subjects) examined cocoa, chocolate, or flavan-3-ols and had relevant outcomes. The studies variously included subjects who were healthy, overweight, hypertensive, hypercholesterolemic (high cholesterol levels), diabetic, and subjects who had elevated CVD risk, congestive heart failure, stable coronary artery disease, smoking-related endothelial (inner arterial tissue) dysfunction, chronic fatigue, or combined hypertension and impaired glucose tolerance. They consisted of 15 parallel-design and 26 crossover trials.

Interventions included cocoa drinks (21 trials), dark or milk chocolate (15 trials), cocoa supplements (3 trials), solid chocolate plus cocoa drinks (2 trials), and a whole diet (all foods provided) including cocoa powder and chocolate (1 trial). The controls were with low flavan-3-ol versions of the same foods, drinks, or supplements and were fairly well controlled in 23 studies. The placebos were unclear in 4 studies.

All studies were at medium-to-high risk of bias due to poor reporting of allocation concealment, blinding, dropouts, and use of commercial funding, according to the authors. Allocation concealment was adequate in 10 trials, inadequate in 1 trial, and unclear in 31 trials. Participant blinding was adequate in 24 trials, inadequate in 13 trials, and unclear in 5 trials. Provider/researcher blinding and outcome-assessor blinding were adequate in approximately 50% of the studies and were unclear in the rest. There was no reported funding from commercial companies in 8 studies and commercial funding reported in 28 studies, while 6 were unclear with regard to funding. The assessed percentage of energy from saturated fat in the intervention group was 2% or more of that in the control group in 8 studies, and 2% or more in 6 studies (suggesting significant dissimilarity of diet between intervention and control groups that may have impacted outcomes); the remainder of the studies did not report on fat intake.

The meta-analysis suggested improvement in FMD both acutely (2 hours after ingestion of chocolate/cocoa; 3.19%; 95% confidence interval [CI]: 2.04%, 4.33%; 11 studies, 373 participants, I2=84%) and after chronic intake (1.34%; 95% CI: 1.00%, 1.68%; 11 studies, 382 participants, I2=0%). There was also a significant reduction in fasting serum insulin concentrations (-2.65 µU/mL), serum insulin after glucose challenge (-17 µU/mL; 95% CI: -20.7, -13.4 µU/mL), and homeostasis model assessment-insulin resistance (HOMA-IR; -0.67; 2 trials, 70 participants, I2=60%) after chocolate or cocoa interventions. The combined effect of reduced HOMA-IR with improved FMD could be substantial for reducing cardiovascular risk. There was no effect on fasting glucose, HbA1c, or triglycerides.

Significant reductions in diastolic BP (-1.60 mmHg; 95% CI: -2.77, -0.43 mmHg; 22 trials, 918 participants, I2=52%) and mean arterial pressure (-1.64 mmHg; 95% CI: -3.27, -0.01; 4 trials, 163 participants, I2=0%) after chronic intake were observed. Marginally significant effects on LDL (-0.07 mmol/L; 95% CI: -0.14, -0.00 mmol/L; 21 studies, 986 participants, I2=58%) and HDL (0.03 mmol/L; 95% CI: 0.00, 0.06 mmol/L; 21 studies, 986 participants, I2=67%) cholesterol were found. There were no significant effects on diastolic BP after acute intake, or on CRP, total cholesterol, or systolic BP after acute or chronic intake. Too few trials reported on body weight, body mass index (BMI), and waist circumference to do an analysis.

The effects of other factors, such as sex, dose, duration, etc., were examined where there were enough studies reporting on those outcomes. There were significant improvements in FMD for all doses of epicatechin (an antioxidant flavanol compound found in cocoa and chocolate ingredients), for either short or long durations, with greater improvements at higher doses. There were also improvements with epicatechin for systolic and diastolic BP at doses >50 mg/d, and in fasting glucose and triglycerides at doses of 50-100 mg/d. Only short-term studies (<3 weeks) reduced fasting glucose, LDL, and total cholesterol; and only long-term studies (>3-wks. duration) increased HDL cholesterol.

Removing studies funded by industry did not change the results for FMD or HOMA-IR, but the effects on BP and HDL and LDL cholesterol were no longer significant. Removing studies with unclear allocation concealment also did not change the results for FMD or HOMA-IR, but the short-term chronic effects on diastolic BP and the effects on systolic BP became statistically significant.

Funnel plots showed a minor trend towards bias for FMD, and no bias for systolic BP (though in the latter, the power was limited due to small study size).

According to the authors, this is the first systematic review and meta-analysis to assess the effects and validity of all RCTs on cocoa and chocolate with respect to many important CVD risk factors. It shows (for the first time) that cocoa and chocolate reduce insulin resistance as a result of reduced insulin secretion. It also shows a strong effect on FMD that is of real clinical significance. There was a moderately robust effect on diastolic BP, triglycerides, and mean arterial pressure, and marginally significant effects on LDL and HDL cholesterol. The authors also noted, “Our results support the reciprocal relation between insulin resistance and endothelial function . . . epicatechin dose may be a key contributor to the effects observed.”

—Risa Schulman, PhD