Effect of cocoa products on blood pressure: systematic review and meta-analysis.

Abstract

Cocoa products such as dark chocolate and cocoa beverages may have blood pressure (BP)-lowering properties due to their high content of plant-derived flavanols.
We performed a meta-analysis of randomized controlled trials assessing the antihypertensive effects of flavanol-rich cocoa products. The primary outcome measure was the change in systolic and diastolic BP between intervention and control groups.
In total, 10 randomized controlled trials comprising 297 individuals were included in the analysis. The populations studied were either healthy normotensive adults or patients with prehypertension/stage 1 hypertension. Treatment duration ranged from 2 to 18 weeks. The mean BP change in the active treatment arms across all trials was -4.5 mm Hg (95% confidence interval (CI), -5.9 to -3.2, P < 0.001) for systolic BP and -2.5 mm Hg (95% CI, -3.9 to -1.2, P < 0.001) for diastolic BP.
The meta-analysis confirms the BP-lowering capacity of flavanol-rich cocoa products in a larger set of trials than previously reported. However, significant statistical heterogeneity across studies could be found, and questions such as the most appropriate dose and the long-term side effect profile warrant further investigation before cocoa products can be recommended as a treatment option in hypertension.

Full text

Page 1

AMERICAN JOURNAL OF HYPERTENSION | VOLUME 23 NUMBER 1 | 97-103 | jaNUaRy 2010 97
original contributions
nature publishing group
Cocoa products such as dark chocolate and cocoa beverages
may have blood pressure (BP)–lowering properties due to their
high content of plant-derived flavanols.1 Initial observations of
the possible antihypertensive effects of cocoa-rich food came
from the Kuna Indians, a native population living on islands
off the Panama coast. Hypertension is very rare in the island
Kuna Indians, and there is no age-dependent increase in BP.
These effects are likely environmental because they are lost
upon migration to urban Panama City. One striking dietary
feature of the Kuna Indians living offshore is the consumption
of large amounts of natural cocoa drinks rich in flavanols.2 In
contrast, Kuna Indians living in Panama City consume cocoa
from grocery stores largely devoid of flavanols leading to the
hypothesis that intake of flavanol-rich cocoa might be causally
linked to the low prevalence of hypertension in island- dwelling
Kuna Indians.
Further epidemiological evidence comes from a cohort
of 470 men from the Dutch Zutphen Elderly Study. Habitual
consumption of cocoa-containing food was inversely
associated with BP.3 Mean systolic and diastolic BP were 3.7
and 2.1 mm Hg lower in participants in the highest tertile
of cocoa intake as compared to those in the lowest tertile.
Furthermore, cocoa intake was inversely related to cardiovas-
cular and all-cause mortality.
Sparked by these epidemiological observations, several ran-
domized clinical trials examining the effect of cocoa-rich prod-
ucts on BP have been undertaken. In 2007, a meta- analysis
of randomized studies examining the effect of cocoa- and
flavanol-rich dark chocolate consumption on BP reported
beneficial effects as compared to chocolate containing no or
only negligible amounts of flavanols.4 However, only five small
trials were included in the analysis with a total of 96 partici-
pants. Several randomized controlled trials have been pub-
lished since suggesting an updated meta-analysis with a more
reliable assessment of the antihypertensive effects of flavanol-
rich cocoa products.
Methods
Literature search. We performed a search of MEDLINE via
PubMed (1966 to June 2009), EMBASE (1980 to June 2009),
CENTRAL (The Cochrane Controlled Clinical Trials Register)
and the ClinicalTrials.gov Web site to identify randomized
controlled trials examining the effect of dark chocolate or
1Department of Cardiology, University of Leipzig–Heart Center, Leipzig,
Germany; 2Clinical Trial Service Unit, University of Oxford, Oxford, UK.
Correspondence: Steffen Desch (stdesch@web.de)
Received 15 September 2009; first decision 11 October 2009; accepted 14 October
2009; advance online publication 12 November 2009. doi:10.1038/ajh.2009.213
© 2010 American Journal of Hypertension, Ltd.
Effect of Cocoa Products on Blood Pressure:
Systematic Review and Meta-Analysis
Steffen Desch1, Johanna Schmidt1, Daniela Kobler1, Melanie Sonnabend1, Ingo Eitel1, Mahdi Sareban1,
Kazem Rahimi2, Gerhard Schuler1 and Holger Thiele1
Background
Cocoa products such as dark chocolate and cocoa beverages may
have blood pressure (BP)–lowering properties due to their high
content of plant-derived flavanols.
Methods
We performed a meta-analysis of randomized controlled trials
assessing the antihypertensive effects of flavanol-rich cocoa
products. The primary outcome measure was the change in systolic
and diastolic BP between intervention and control groups.
results
In total, 10 randomized controlled trials comprising 297 individuals
were included in the analysis. The populations studied were either
healthy normotensive adults or patients with prehypertension/
stage 1 hypertension. Treatment duration ranged from 2 to 18 weeks.
The mean BP change in the active treatment arms across all trials
was −4.5 mm Hg (95% confidence interval (CI), −5.9 to −3.2, P < 0.001)
for systolic BP and −2.5 mm Hg (95% CI, −3.9 to −1.2, P < 0.001) for
diastolic BP.
conclusions
The meta-analysis confirms the BP-lowering capacity of flavanol-
rich cocoa products in a larger set of trials than previously reported.
However, significant statistical heterogeneity across studies could
be found, and questions such as the most appropriate dose and the
long-term side effect profile warrant further investigation before
cocoa products can be recommended as a treatment option in
hypertension.
Keywords: arterial hypertension; blood pressure; clinical trials; cocoa;
hypertension; meta-analysis
Am J Hypertens 2010; 23:97-103 © 2010 American Journal of Hypertension, Ltd.

Page 2

98
jaNUaRy 2010 | VOLUME 23 NUMBER 1 | AMERICAN JOURNAL OF HYPERTENSION
original contributions
Cocoa and Blood Pressure
cocoa-containing beverages on arterial BP. Furthermore, we
hand-searched bibliographic citations from the retrieved
papers and from review articles. No language restrictions were
applied.
MEDLINE and EMBASE were searched using the term
“cacao” (major subject heading) OR the text words “cacao”
OR “cocoa” OR “chocolate.” The search was limited to the
criteria “clinical trials” and “human.” CENTRAL and the
ClinicalTrials.gov Web site were searched with “chocolate OR
cocoa OR cacao” as free terms without further restrictions
applied to the search.
Two investigators (S.D. and J.S.) independently screened
the titles and abstracts resulting from the search strategies.
Articles were rejected on initial screening if titles or abstracts
were clearly irrelevant. The full text of potentially relevant arti-
cles was reviewed to assess eligibility for the inclusion in the
meta-analysis with any disagreement resolved by consensus.
Selection of studies. Study design had to meet the following
criteria: (i) investigating flavanol-rich cocoa products such as
dark chocolate and cocoa beverages; (ii) random allocation to
treatment and control group; (iii) BP measurements at baseline
and at a minimum of one more time point; (iv) because we
wanted to study the effects of habitual intake of cocoa products
on BP, single-dose trials were not included in the analysis, and
a minimum of 2 weeks of treatment duration was required. No
restrictions were made with regard to age, gender, medication,
baseline BP, risk profile, or comorbidities. None of the indi-
viduals enrolled took any antihypertensive medication.
Data extraction. Data were independently extracted by two
reviewers (S.D. and J.S.) using a standard form and cross-
checked. When BP measurement data were available for both
ambulatory BP monitoring and office BP, preference was given
to values obtained by ABPM. The position of the patient dur-
ing BP measurement may affect the BP-lowering effect. When
BP measurement data were available in more than one posi-
tion, data were extracted with the following order of prefer-
ence: (i) sitting; (ii) standing; (iii) supine. If only one position
was reported, data from that position were used.
Assessment of trial quality and risk of bias. Two independent
reviewers (S.D. and J.S.) assessed the quality as well as the risk
of bias of the trials included and created a risk of bias table
according to recommendations by the Cochrane Collaboration.5
We did not create funnel plots because the limited number of
trials included and the lack of studies with large sample sizes
preclude a meaningful interpretation.
Statistical analysis. The change in systolic and diastolic BP
between the intervention and control groups was defined as
the primary outcome measure. For parallel-group trials, the
treatment effect was defined as the mean difference of the
treatments (change from baseline in BP) between groups.
None of the four parallel-group trials included in the meta-
analysis directly reported standard deviations (SD) for the
mean difference of the treatments. In one trial, the SD could be
calculated from confidence intervals (CIs).6 In the other three
parallel-group trials, we estimated the SD based on the SD of
BP at the end of treatment. For crossover trials, the treatment
effect was defined as the mean within-subject difference of the
treatments (change from baseline in BP) assuming no carry-
over or period effect. To properly analyze crossover trials, it is
necessary that the mean difference of the treatments and the
corresponding standard error (which can be calculated from
SD or CIs, and sample size) are available.7–9 In one trial, the
standard error could be calculated from the sample size and
the SD of the mean difference of the treatments.10 In another
trial, the standard error was estimated from a P value for the
mean difference of the treatments.11 If the standard error for
the within-person differences could not be extracted, correla-
tion between treatment outcomes was approximated using a
conservative estimate of the correlation coefficient (r = 0.68)
derived from two trials in which individual pre- and post-
treatment BP values were available.4 Thereafter, the corre-
sponding standard error was calculated as described in the
Cochrane Handbook for Systematic Reviews of Interventions.7
Statistical heterogeneity between the trials was tested using a
standard χ2 statistic with a P value set at 0.10. The I2 statistic
was used to examine inconsistencies across studies. To com-
bine both parallel-group and crossover trials in the final
analysis, the generic inverse variance method was used and a
random effects model applied.7 Sensitivity analyses were per-
formed to test the robustness of the results. Subgroup analyses
were carried out by comparing treatment effects (i) between
the seven short-term trials10–16 with a treatment duration of 2
weeks vs. the three medium-term trials6,17,18 with a treatment
duration between 4 and 18 weeks; (ii) between studies with
lower12,13,15,17,18 and higher6,10,11,14,16 baseline BPs (groups
divided by the median of baseline systolic BP); (iii) between
studies with lower6,12–15 and higher10,11,16–18 flavanol content
in the active arm (groups divided by the median of flavanol
332 Articles identified from literature
search
292 Articles judged irrelevant by title or
abstract
30 Articles excluded
• Study design did not meet
prespecified criteria (e.g., single-dose
trial, control group not applicable,
nonrandomized trial), n = 16
• Blood pressure reporting insufficient
to calculate effect size, n = 1
• Publication/study did not examine
blood pressure, n = 13
40 Potentially relevant articles
identified for full text review
10 Randomized controlled trials
identified
Figure 1 | Flowchart of the study selection process.

Page 3

AMERICAN JOURNAL OF HYPERTENSION | VOLUME 23 NUMBER 1 | jaNUaRy 2010 99
original contributions
Cocoa and Blood Pressure
content). Data synthesis and statistical analyses were done
using the Cochrane Collaboration Review Manager (RevMan,
version 5.0.18; The Cochrane Collaboration, Copenhagen,
Denmark). The study was performed in compliance with the
Quality of Reporting of Meta-analyses guidelines.19
results
The study selection process is shown in Figure 1. Among
the trials excluded from the analysis, two studies compar-
ing low-flavanol vs. high-flavanol beverages could not be
considered20,21 because the amount of flavanols ingested in the
low-flavanol control arm was significantly higher than those
reported to possess significant BP-lowering capacity.6 Another
study reported mean pressures only and was therefore not
included in the analysis.22 In total, 10 randomized control-
led trials examining the effect of cocoa products on BP were
included in the analysis6,10–18 (Table 1). A crossover design
was used in six studies and a parallel-group design in four
studies. Combining all studies, 297 individuals were analyzed.
The populations studied were either healthy normotensive
adults or patients with prehypertension/stage 1 hypertension
without antihypertensive medication. The majority of stud-
ies used office BP to assess treatment effects. Treatment dura-
tion ranged from 2 to 18 weeks. Flavanol intake varied widely
across studies, e.g., between 5 and 174 mg for the flavanol sub-
compound epicatechin (Table 1).
Results of the quality assessment are summarized in Figure 2.
Among the trials using a “dark vs. white chocolate” design (as
opposed to cocoa beverages or tablets), adequate blinding of
participants was not possible due to the obvious differences
in appearance and taste. However, investigators and end point
assessment were blinded in most studies. Other minor sources
of potential bias included the following: the majority of trials (i)
did not describe the methods of sequence generation in the ran-
domization process, (ii) did not provide sufficient data to per-
mit judgment if participants or investigators could foresee the
assignment to the treatment groups (allocation concealment),
(iii) provided insufficient data to rule out selective reporting.
The mean BP reduction in the active treatment arms across
all trials was −4.5 mm Hg (95% CI, −5.9 to −3.2, P < 0.001) for
systolic BP and −2.5 mm Hg (95% CI, −3.9 to −1.2, P < 0.001)
for diastolic BP (Figure 3a,b). Significant statistical heteroge-
neity across studies could be found in the initial analysis for
both systolic BP (χ2 = 84, P < 0.01; I2 = 89%) and diastolic BP
(χ2 = 92, P < 0.01; I² = 90%). Sensitivity analyses revealed that
three studies were partly responsible.6,13,14 Two of these stud-
ies reported exceptionally large reductions in BP.13,14 When
excluding all three studies from the analysis, the I2 statistic
was reduced from 89 to 43% (χ2 = 10; P = 0.11) for systolic BP.
The pooled reduction in systolic BP changed only slightly by
Study
a
b
Taubert 2003 (ref. 10)
Murphy 2003 (ref. 17)
Engler 2004 (ref. 12)
Grassi 2005 (ref. 13)
Fraga 2005 (ref. 15)
Grassi 2005 (ref. 14)
Taubert 2007 (ref. 6)
Grassi 2008 (ref. 16)
Crews 2008 (ref. 18)
Muniyappa 2008 (ref. 11)
Crossover: 13
Cocoa: 13; Control: 15
Cocoa: 11; Control: 10
Crossover: 15
Crossover: 27
Crossover: 20
Cocoa: 22; Control: 22
Crossover: 19
Cocoa: 45; Control: 45
Crossover: 20
13.1%
2.7%
9.7%
13.3%
14.5%
11.2%
14.6%
12.2%
4.7%
4.0%
−10
Favors cocoa
0
Favors control
510
−5
−5.10 [−6.40, −3.80]
1.00 [−6.43, 8.43]
−1.80 [−4.37, 0.77]
−6.20 [−7.41, −4.99]
−4.00 [−4.59, −3.41]
−11.00 [−13.02, −8.98]
−3.00 [−3.50, −2.50]
−4.60 [−6.26, −2.94]
−0.53 [−5.69, 4.63]
−1.00 [−6.82, 4.82]
Total (95% CI) 297100.0%
−4.52 [−5.87, −3.16]
Sample sizeWeightChange in systolic BP (mm Hg), 95% CI
Study
Taubert 2003 (ref. 10)
Murphy 2003 (ref. 17)
Engler 2004 (ref. 12)
Grassi 2005 (ref. 13)
Fraga 2005 (ref. 15)
Grassi 2005 (ref. 14)
Taubert 2007 (ref. 6)
Grassi 2008 (ref. 16)
Crews 2008 (ref. 18)
Muniyappa 2008 (ref. 11)
11.8%
3.7%
10.7%
12.0%
12.5%
10.5%
12.3%
10.5%
7.6%
8.5%
−10
Favors cocoa
0
Favors control
510
−5
−1.80 [−2.89, −0.71]
−1.00 [−6.94, 4.94]
1.00 [−0.71, 2.71]
−3.50 [−4.51, −2.49]
−4.00 [−4.59, −3.41]
−8.30 [−10.10, −6.50]
−1.90 [−2.70, −1.10]
−4.20 [−6.00, −2.40]
0.07 [−3.07, 3.21]
1.00 [−1.72, 3.72]
Total (95% CI) 297100.0%
−2.52 [−3.87, −1.16]
Crossover: 13
Cocoa: 13; Control: 15
Cocoa: 11; Control: 10
Crossover: 15
Crossover: 27
Crossover: 20
Cocoa: 22; Control: 22
Crossover: 19
Cocoa: 45; Control: 45
Crossover: 20
Sample size Weight Change in diastolic BP (mm Hg), 95% CI
Figure 3 | Effect of cocoa product intake on (a) systolic blood pressure and (b) diastolic blood pressure. BP, blood pressure; CI, confidence interval.
Adequate sequence generation?
Allocation concealment?
Blinding?
Incomplete outcome data addressed?
Free of selective reporting?
Free of other bias?*
0%
Yes (low risk of bias)No (high risk of bias) Unclear
25% 50%75%100%
Figure 2 | Risk of bias graph. *e.g., the trial had extreme baseline imbalances;
was stopped early; etc.

Page 4

100
jaNUaRy 2010 | VOLUME 23 NUMBER 1 | AMERICAN JOURNAL OF HYPERTENSION
original contributions
Cocoa and Blood Pressure
table 1 | characteristics of trials included
Study
Taubert 2003
(ref. 10)
Murphy 2003
(ref. 17)
Engler 2004
(ref. 12)
Grassi 2005
(ref. 13)
Fraga 2005
(ref. 15)
Grassi 2005
(ref. 14)
Taubert 2007
(ref. 6)
Grassi 2008 (ref.
16)
Crews 2008
(ref. 18)
Muniyappa 2008
(ref. 11)
Intervention
Flavanol-rich
dark chocolate
vs. flavanol-free
white chocolate
Flavanol-rich
tablets vs.
flavanol-poor
tablets
Flavanol-rich
dark chocolate
vs. flavanol-
free white
chocolate
Flavanol-rich
dark chocolate
vs. flavanol-free
white chocolate
Flavanol-rich
milk chocolate
vs. flavanol-poor
chocolate
Flavanol-rich
dark chocolate
vs. flavanol-free
white chocolate
Flavanol-rich
dark chocolate
vs. flavanol-free
white chocolate
Flavanol-rich
dark chocolate
vs. flavanol-free
white chocolate
Flavanol-rich
dark chocolate
and flavanol-rich
cocoa beverage
vs. matching
placebo
Flavanol-rich
cocoa drink vs.
flavanol-poor
cocoa drink
Flavanol intake
in intervention
group
Catechin
27 mg/d
Epicatechin
81 mg/d
234 mg Cocoa,
flavanols, and
procyanidins
(catechin and
epicatechin
values not stated)
46 mg
Epicatechin/d
213 mg
Procyanidins/d
168 mg
Flavanols/d
22 mg
Catechin
66 mg
Epicatechin
168 mg
Flavanols/d
39 mg Catechin
and epicatechin
126 mg
Procyanidins
168 mg
Flavanols/d
22 mg Catechin 66 mg
Epicatechin
Catechin
1.7 mg/d
Epicatechin
5.1 mg/d
Catechin
36 mg/d
Epicatechin
111 mg/d
754 mg
Procyanidins/d
(catechin and
epicatechin
values
not stated)
902 mg
Flavanols/d
Catechin 62 mg/d
Epicatechin
174 mg/d
Procyanidins
676 mg/d
Flavanol intake
in control group
Negligible
Negligible
Negligible
Negligible
Negligible
Negligible
Negligible
Negligible
Negligible
Negligible
Design
Randomized,
crossover,
14 days
Randomized,
parallel-group,
28 days
Randomized,
parallel-group,
controlled,
2 weeks
Randomized,
crossover,
15 days
Randomized,
crossover,
14 days
Randomized,
crossover,
15 days
Randomized,
parallel-group,
18 weeks
Randomized,
crossover,
15 days
Randomized,
parallel-group,
6 weeks
Randomized,
crossover,
2 weeks
Population
Patients with
isolated systolic
hypertension
(grade 1)
Healthy
normotensive
adults
Healthy
normotensive
adults
Healthy
normotensive
adults
Healthy
normotensive
young male
adults
Patients with
hypertension
(grade 1)
Patients with
prehypertension
or hypertension
(grade 1)
Patients with
hypertension
(grade 1) and
impaired glucose
tolerance
Healthy, older
adults
Patients with
hypertension
(grade 1)
Number of
participants
13
32
(28 analyzed)
22
(21 analyzed)
15
28
(27 analyzed)
20
44
19
101
(90 analyzed)
29
(20 analyzed)
Missing
participants
0
4
1
0
1
0
0
0
11
9
table 1 | continued on next page

Page 5

AMERICAN JOURNAL OF HYPERTENSION | VOLUME 23 NUMBER 1 | jaNUaRy 2010 101
original contributions
Cocoa and Blood Pressure
Men/women (n) 7/6
17/15
11/11
7/8
28/0
10/10
20/24
11/8
41/60
8/12
age (years)
58.8 (mean),
range 55–64
active group:
40 ± 9
Placebo group:
47 ± 4
(means ± s.d.)
Range 21–55
33.9 ± 7.6
(mean ± s.d.)
Range 18–20
43.7 ± 7.8
(mean ± s.d.)
DC group:
63.4 ± 4.7
WC group:
63.7 ± 4.8
(means ± s.d.)
44.8 ± 8.0
(mean ± s.d.)
Chocolate and
cocoa group:
68.8 ± 8.6
Placebo group:
68.7 ± 8.0
(means ± s.d.)
51 ± 1.5
(mean ± s.e.)
Baseline BP
intervention
group,
systolic ± s.d./
diastolic ± s.d.
(mm Hg)
153 ± 4/85 ± 5
118 ± 13/78 ± 12 121 ± 5/68 ± 3
109 ± 8/72 ± 5
123 ± 3/72 ± 2
136 ± 6/88 ± 4
148 ± 7/86 ± 4
135 ± 4/87 ± 4
127 ± 14/74 ± 8
141 ± 3/91 ± 3
Baseline BP
control group,
systolic ± s.d./
diastolic ± s.d.
(mm Hg)
154 ± 4/84 ± 4
116 ± 9/76 ± 8
112 ± 8/66 ± 2
110 ± 8/72 ± 5
123 ± 3/71 ± 2
136 ± 6/88 ± 4
148 ± 8/87 ± 4
134 ± 4/87 ± 4
129 ± 14/75 ± 8
140 ± 2/87 ± 2
antihypertensive
medication
None
None
None
None
None
None
None
None
None
None
Details of BP
measurement
Office BP
(automated
oscillometric
device), seated
position
Office BP
(automated
oscillometric
device),
no data on
body position
Office BP
(automated
oscillometric
device), seated
position
Office BP,
sphygmomano-
meter, seated
position
Office BP
(automated
oscillometric
device),
no data on
body position
aBPM
Office BP
(automated
oscillometric
device), seated
position
aBPM
Office BP
(automated
oscillometric
device), seated
position
Office BP,
sphygmomano-
meter, seated
position
ABPM, ambulatory blood pressure monitoring; BP, blood pressure.
table 1 | continued
Study
Taubert 2003
(ref. 10)
Murphy 2003
(ref. 17)
Engler 2004
(ref. 12)
Grassi 2005
(ref. 13)
Fraga 2005
(ref. 15)
Grassi 2005
(ref. 14)
Taubert 2007
(ref. 6)
Grassi 2008 (ref.
16)
Crews 2008
(ref. 18)
Muniyappa 2008
(ref. 11)