Impact of cocoa flavanols on blood pressure and vascular function: an updated systematic review and meta-analysis of randomized controlled trials

Graphical abstract

Abstract

Background and objective

While elevated blood pressure is a major cardiovascular risk factor, a comprehensive evaluation of not only blood pressure but also vascular function is essential to evaluate the risk of cardiovascular diseases and the effectiveness of interventions. Cocoa flavanols may lower blood pressure by improving endothelial function, however evidence regarding vascular indices is limited, and findings remain inconsistent and not the latest. We conducted an updated systematic review and meta-analysis to comprehensively evaluate the effects of cocoa flavanol intake on blood pressure and vascular function.

Methods

This review was registered in UMIN-CTR (UMIN000059995), and we searched PubMed, Cochrane Library, Ichushi-Web, JDreamIII, ICTRP and UMIN-CTR for eligible randomized controlled trials up to December 8, 2025. Primary outcomes were systolic and diastolic blood pressure (SBP/DBP); secondary outcomes were flow-mediated dilation (FMD), pulse wave velocity (PWV), and ankle–brachial index (ABI). Mean differences (MD) with 95% confidence intervals (CIs) were pooled using a random-effects model. Risk of bias was assessed using a modified version of the Cochrane Risk of Bias 2 (RoB 2), and certainty of evidence was graded.

Results

Twenty-five studies (n = 1,167) included hypertensive and normotensive participants; daily cocoa flavanol doses ranged from 10.6 to 1,680 mg. Both SBP and DBP decreased significantly following cocoa flavanol intake (SBP: MD −2.734 mmHg, 95% CI [−3.446, −2.023]; DBP: MD −1.848 mmHg, 95% CI [−2.725, −0.971]). Significant improvements were also observed for FMD (MD 1.314%, 95% CI [1.043, 1.585]) and PWV (MD −0.200 m/s, 95% CI [−0.393, −0.007]), whereas ABI showed no significant change, with limited evidence from only two studies. Subgroup analyses were consistent for SBP/DBP in both hypertensive and normotensive participants. Dose–response analyses revealed a significant inverse association between SBP or DBP and flavanol intake, but did not identify clear dose thresholds within the examined intake range. Certainty was moderate for SBP; other outcomes were limited.

Conclusion

These findings suggest that cocoa flavanol intake of 10.6–1,680 mg/day may modestly reduce blood pressure and may improve selected vascular function indices, although certainty for vascular outcomes remains limited.

1. Introduction

Cardiovascular disease (CVD) is the leading cause of death worldwide [1]. With population aging contributing substantially to its burden, preventing CVD through interventions targeting modifiable risk factors is a key strategy [2]. In 2022, approximately 19.8 million people—approximately 32% of all deaths—died from heart attacks or strokes [1]. Elevated systolic blood pressure (SBP) has been shown to contribute most to mortality risk [2,3]. In later life, age-related vascular alterations—often described as vascular aging—contribute to hypertension, arterial stiffening, and end-organ vulnerability [4]. Therefore, effective management of blood pressure is crucial for the primary prevention of CVD.

Moreover, hypertension is closely related to structural and functional vascular changes, such as endothelial dysfunction and increased arterial stiffness [5]. Blood pressure is clinically important; however, it represents a hemodynamic consequence and does not directly capture the underlying vascular structural and functional alterations that progress with aging [4,6]. Recently, the significance of indices for evaluating vascular function and blood pressure has been highlighted [5]. Flow-mediated vasodilation (FMD) is widely used as an index of endothelial function [7], pulse wave velocity (PWV) as an index of arterial stiffness [5], and the ankle-brachial index (ABI) as an index of peripheral arterial disease and lower limb blood flow [8]. Consequently, in order to estimate the risk of developing CVD and the impact of intervention for CVD, it is now important to comprehensively evaluate not only blood pressure but also structural and functional vascular changes, which cannot be detected by blood pressure alone [5,9].

While pharmacotherapy is widely used to control blood pressure, nonpharmacological approaches, especially dietary interventions, represent an important adjunct [10,11]. Daily consumption of a healthy diet can prevent elevated blood pressure, which is better for the long-term prognosis than starting a therapy after the diagnosis of hypertension [10,11]. However, sustained behavioral changes are often difficult to maintain, and excessive dietary restrictions may impair quality of life [[12], [13], [14]]. Thus, effective dietary interventions without major lifestyle changes are required [12,13].

Bioactive dietary components may play an important role in supporting dietary interventions for the prevention and management of CVD. Polyphenols are commonly studied as bioactive substances, and foods rich in polyphenols may have beneficial effects on vascular function [12], in which they promote vasodilation by improving vascular endothelial function via activation of the nitric oxide (NO) pathway [15]. From this perspective, foods rich in bioactive compounds such as polyphenols may help support dietary interventions aimed at cardiovascular health. In particular, dark chocolate contains a high content of cocoa-derived flavanols, allowing us to incorporate polyphenols into our diet without increasing usual food volume—an aspect that may be particularly relevant for older adults, who often tend to eat less overall and have difficulty chewing and swallowing [16].

As new evidence on the effects of cocoa flavanols on blood pressure and vascular function has emerged since earlier meta-analyses, an updated synthesis of the available evidence is warranted. However, these results about the effects of cocoa flavanol intake on blood pressure and vascular function have been inconsistent [17]. Moreover, previous meta-analyses have mainly reported on blood pressure and/or endothelial function, and comprehensive assessments incorporating multiple indices of vascular function, such as PWV and ABI, have been limited [18]. Although a recent meta-analysis has reported modest reductions in blood pressure and improvements in endothelial function following interventions rich in flavan-3-ols [12], this study evaluated heterogeneous flavanol sources, making it difficult to interpret the evidence as being specifically limited to cocoa-derived flavanols. Furthermore, the previous meta-analyses focused on cocoa-flavanols have primarily assessed blood pressure or cardiometabolic markers, but have not quantitatively integrated multiple vascular function indices together with blood pressure [19,20].

Therefore, we conducted an updated and comprehensive systematic review and meta-analysis to evaluate the effects of sustained cocoa flavanol intake on blood pressure and multiple vascular function indices such as FMD, PWV, and ABI. In addition, subgroup analyses for blood pressure and exploratory dose–response analyses were performed. By quantitatively synthesizing both hemodynamic and vascular functional outcomes using evidence identified through December 8, 2025, this study aims to clarify the potential role of cocoa flavanols in cardiovascular health.

2. Methods

This systematic review and meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) 2020 statement [21]. This protocol was prospectively developed and registered with the University Hospital Medical Information Network Clinical Trials Registry (UMIN-CTR) (registration number: UMIN000059995). No major methodological changes were made from the registered protocol.

2.1. Search strategy

We systematically searched PubMed, the Cochrane Library, Ichushi‑Web, and JDream III (JSTPlus, JMEDPlus, and JST7580) from database inception to December 8, 2025. In addition, we searched the International Clinical Trials Registry Platform (ICTRP) and UMIN‑CTR to identify ongoing or unpublished studies. Searches were limited to articles published in English or Japanese.

The search strategy was developed to identify randomized controlled trials evaluating the effects of cocoa flavanol‑containing foods or products in adult populations. Because cocoa flavanols have been investigated across a broad range of health outcomes, an unrestricted search would have yielded an unfeasibly large body of literature for screening. Given that the primary objective of this review was to evaluate blood pressure–related outcomes, a limited set of blood pressure–related terms were included in the search strategy as a pragmatic methodological trade‑off to ensure feasibility while maintaining relevance to the review question.

Importantly, outcomes were not used exclusively as eligibility criteria during the search. The presence of blood pressure or vascular function outcomes was confirmed during title/abstract screening, full‑text assessment, and data extraction, in accordance with the prespecified review protocol. To mitigate the risk of missing relevant studies, trial registry searches were also conducted to complement the database search.

Search results were screened after duplicate removal. Detailed search strategies for each database, including search dates and the number of records retrieved, are provided in Supplementary Table S1. Hand‑searching of journals and reference list screening were not performed.

2.2. Eligibility criteria

The eligibility criteria were defined according to PICOS (Population, Intervention, Comparator, Outcomes, Study design). Eligible participants were adults aged ≥18 years, with or without medical conditions. Pregnant and lactating women were excluded from the study.

The intervention was consumption of foods containing cocoa-derived flavanols (primarily (−)-epicatechin, (+)-catechin and their oligomeric procyanidins). Studies were included if the daily intake of cocoa flavanols was quantitatively reported (including total polyphenols) or could be calculated from published information. These criteria applied to eligibility for the main meta-analyses; additional requirements for quantitative dose–response analyses are described separately below. Studies were excluded if the amount or composition of intake was unclear, intake fluctuated during the intervention period, or the effects of cocoa flavanols could not be isolated because of co-interventions that may affect blood pressure. Interventions in which a single ingredient, such as isolated and purified (−)-epicatechin, was consumed alone were excluded because they were different from the interventions in this review (consumption of flavanol-containing foods). The food shape was not restricted.

Comparators included placebo foods containing no or small amounts of cocoa flavanols or no consumption, where “small amounts” is defined as ≤10% of the intervention dose because there is no established standard for a low-flavanol comparator.

Primary outcomes were SBP and diastolic blood pressure (DBP), secondary outcomes included FMD, PWV, and ABI as vascular function indices [5,7,8]. Studies were excluded if they did not measure the primary outcome before or after (or during) the intake period, if data on the mean and variance indices of blood pressure were unavailable, or if the number of participants could not be confirmed.

Eligible study designs were RCTs (parallel-group or crossover). Studies were included if the intervention duration was ≥2 weeks, duration was the same in the intervention and control groups, and the primary outcome data enabled comparison between groups. The minimum intervention duration of 2 weeks was selected to evaluate the sustained effects of cocoa flavanol consumption on blood pressure and vascular function, consistent with previous systematic reviews of cocoa flavanol interventions [18]. Studies were excluded if they reported only the results of the stratified analyses without presenting the results for the overall population. When multiple reports from the same study were found, the information was collated to avoid duplication, and the report with the most complete data on the primary outcome was selected. Studies that combined dietary or exercise interventions that may affect blood pressure, crossover trials without an appropriate washout period, reviews, commentaries, articles that cited the results of original publications, and secondary analyses were excluded. Exercise was defined as supervised exercise (e.g., structured exercise programs or therapy).

2.3. Study selection

Two reviewers independently performed study selection. After removing duplicates, a primary screening based on titles and abstracts was performed, and full-text articles were checked if it was difficult to determine eligibility. Secondary screening based on full-text articles was then performed, and discrepancies were resolved through consultation. Disagreements during consultation were judged by a third reviewer. Automated tools were not used for article selection.

2.4. Data extraction

Data extraction was performed independently by two reviewers. The following items were extracted: author, title, publication year, journal, study design, participant characteristics, intervention and control details, intervention duration, number of participants, and outcome measurements.

For outcomes, the mean and standard deviation values for SBP, DBP, FMD, PWV, and ABI before and after intervention (or mean changes with corresponding variance measures) were extracted. If multiple blood pressure measurement methods were reported, peripheral (office/clinical) blood pressure was prioritized because it was the most consistently reported measure across the included studies. Ambulatory blood pressure monitoring (ABPM) was used only when peripheral blood pressure was unavailable. When both 24-h and daytime/nighttime values were reported, daytime values were selected because they were considered more comparable to the office/clinical blood pressure measurements reported in the majority of included studies. For PWV, carotid–femoral PWV (cfPWV) was prioritized because it is considered the reference standard measure of arterial stiffness [5]. When cfPWV was unavailable, other measures reflecting central arterial stiffness (e.g., aortic PWV) were selected preferentially. Other PWV measures, including heart–ankle PWV (haPWV) and brachial–ankle PWV (baPWV), were used only when cfPWV or other central PWV measures were unavailable. A sensitivity analysis excluding studies contributing only haPWV or baPWV data was conducted. When ABI values were reported separately for the right and left sides, their mean was used to avoid double counting. Interim measurements during the intervention period were excluded. Data extraction was based on information publicly available in published articles and supplementary materials, and missing data from primary studies were not followed up by contacting the authors. This approach was intended to ensure consistency and objectivity in data extraction and to avoid potential bias arising from differential availability of additional information through author contact. When essential outcome data could not be extracted or derived using standard formulas, those studies were excluded from the corresponding meta-analysis. No automated tools were used for data extraction. Discrepancies were resolved through review of the original articles and discussion. Information on adverse events and withdrawals due to adverse events was extracted when available.

2.5. Data synthesis

Quantitative analysis was performed using the mean, standard deviation, and sample size for each group. The results of the main analysis of outcomes with a small number of pooled studies (e.g., <5 studies) and sensitivity analyses with a limited number of studies were interpreted as exploratory findings. In studies that reported both the observed value and the change, the observed value was prioritized. When only standard errors or 95% confidence intervals (95% CIs) were reported, they were converted to standard deviations using a standard formula. When an index of variance for the change was not available, it was excluded from quantitative analysis. Since the information required for a paired analysis (e.g., the variance of within-participant differences) was not sufficiently reported for the crossover studies, data by group for all reported periods were used for the pooled data (approximate approach to address insufficient reporting) [22].

2.6. Subgroup analyses

Furthermore, to assess the robustness of the primary outcome (SBP and DBP), subgroup analyses were performed based on the presence of blood pressure and health status of the study population. The study population was stratified into the following three categories: (1) participants with hypertension, (2) normotensive participants with or without other diseases, and (3) normotensive healthy participants. Healthy participants were defined according to the classifications reported by the original authors. When the 95% CI was not reported, it was calculated using the standard deviation (SD) or standard error (SE). The SD/SE was also estimated from the figures when not reported. Differences between subgroups were not statistically tested as these analyses were exploratory, and the number of studies within each subgroup was limited. These considerations applied to subgroup analyses of pooled mean differences (MDs); dose–response subgroup analyses were conducted and evaluated separately using a dedicated analytical framework. The robustness of the results was evaluated by comparing the results of the analyses with those of the main analysis, and examining whether substantial changes occurred in the direction and magnitude of the effect estimates. Analyses were performed using Comprehensive Meta-Analysis software (version 4.0.0, Biostat, USA).

2.7. Statistical analysis

The mean difference was used for the effect measure because outcomes were assessed on a common continuous scale. Pooled analyses were performed using a random-effects model (inverse variance method) considering the clinical and methodological heterogeneity among the studies. When the same control group was compared with multiple intervention groups, the number of participants in the control group was divided across comparisons to avoid double counting.

Between-study heterogeneity was assessed using the I² statistic, Cochran's Q statistic, and τ² estimates [23]. An I² of ≥50% was used as a threshold to prompt sensitivity analyses exploring potential sources of heterogeneity. Studies whose 95% CIs did not overlap with the pooled 95% CI were considered outliers and were excluded in sensitivity analyses. If there were no such studies, robustness of the pooled estimates was evaluated by excluding studies with the greatest deviation from the pooled 95% CI. Statistical significance was set at p < 0.05.

2.8. Dose-response analysis

Dose–response relationships were examined using the two-stage dose–response analysis approach proposed by Crippa and Orsini [24]. MDs in blood pressure between the intervention group and the control group were calculated for each trial, and variances were derived from reported standard deviations and sample sizes. Total cocoa flavanol intake (mg/day) was used as the exposure variable. When individual flavanol components (e.g., epicatechin and procyanidins) were reported, their summed values were used. Trials reporting only total polyphenol content without quantifiable flavanol data were excluded from the dose–response analyses.

In this study, a linear dose–response model was used to examine dose–response relationship, as non-linear dose–response relationships could not be estimated. Study-specific dose–response slopes were estimated and pooled using generalized least squares under a random-effects model, with between-study variance estimated by restricted maximum likelihood (REML). In addition, dose–response analyses were conducted separately in hypertensive populations and normotensive populations. Their slopes were estimated separately and compared using Wald-type tests as an approximation of interaction.

2.9. Risk of bias assessment

Risk of bias in each included study was assessed for each outcome using a modified version of the Cochrane Risk of Bias 2 (RoB 2) tool. The RoB 2 tool for parallel RCTs and the RoB 2 extension for crossover trials were applied, as appropriate [25].

In parallel-group trials, five domains were evaluated: randomization process, deviation from the intended intervention, missing outcomes, outcome measurement, and selection of reported results. In crossover trials, one additional domain was further evaluated: period effect and carry-over effect. Each domain was judged as “low risk,” “some concerns,” and “high risk,” and these judgements were combined according to the RoB 2 algorithm to determine the overall risk of bias. Assessments were performed independently by two reviewers, and discrepancies were resolved through discussion or adjudication with or by a third reviewer.

2.10. Publication bias assessment

Publication bias was assessed by visual inspection of funnel plot asymmetry for each outcome. Publication bias was statistically examined using Egger’s regression test (p < 0.1) [26], which was performed only for outcomes with ≥3 pooled studies. However, outcomes with <10 pooled studies were interpreted as reference findings because of the limited reliability of Egger’s test under such conditions. If the Egger’s test showed significance, the trim-and-fill method was used to calculate the adjusted pooled effect [27], and the impact of potential publication bias on the results was assessed. For outcomes with <3 pooled studies, funnel plots were constructed using Review Manager software (RevMan, version 10.1.1, The Cochrane Collaboration).

2.11. GRADE assessment

Certainty of evidence was assessed using the GRADE (Grading of Recommendations Assessment, Development and Evaluation) approach. The five domains of the assessment were risk of bias, indirectness, imprecision, inconsistency, and publication bias. Downgrading was performed as necessary based on an overall assessment of each component. For each outcome, the certainty of evidence was graded on a 4-point scale: high, moderate, low, or very low. The assessments were performed independently by two reviewers, and discrepancies were resolved through consultation or by a third reviewer.

3. Results

3.1. Study selection

A total of 1,187 records were identified by searching the databases (PubMed, The Cochrane Library, Ichushi-Web, and JDreamIII) and trial registry databases (ICTRP and UMIN-CTR). After excluding 283 duplicates, 904 records were screened based on titles and abstracts, of which 769 records were excluded. Following full-text assessment of 135 records and exclusion of 110 records, 25 reports were finally included (Fig. 1) [17,[28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48], [49], [50], [51]]. The main reasons for exclusion at full-text assessment were outcome not relevant (SBP/DBP not assessed) (n = 29), intervention duration <2 weeks (n = 20), co-intervention with other potentially BP-active ingredients (n = 20), inappropriate comparator (not placebo or no-intervention) (n = 11), intervention not a cocoa flavanol–containing food (n = 6), SBP/DBP data unavailable for meta-analysis (n = 5), not an RCT (n = 1), pregnant or lactating participants (n = 1), not a peer-reviewed original research article (n = 1), and other reasons (n = 16) (Fig. 1). The list of studies excluded from the full-text evaluation (including the reasons for exclusion) is shown in Supplementary Table S2. The search for unreported studies included 76 clinical trials (Supplementary Table S3), of which eight were relevant to PICO criteria.

Fig. 1.

PRISMA 2020 flow diagram of study selection.

3.2. Study characteristics

Characteristics of the 25 selected studies are presented in Table 1. The study design included 19 parallel-group RCTs and six crossover trials. The primary outcomes were SBP (n = 25) and DBP (n = 24). As a secondary outcome, FMD was evaluated in seven studies, but the data necessary for quantitative integration were available in six studies. PWV and ABI were evaluated in six and two studies, respectively. The intake period ranged from 14 days to 6 months. The daily intake of cocoa flavanols ranged from 10.6 to 1,680 mg.

Table 1.

Characteristics of included studies.

Study	Design	Country	Population / Healthy status	Baseline BP	Age	Duration	Intervention group (n)	Control group (n)	Intervention characteristic	Control characteristic	Outcomes (review-relevant)	Safety assessment
Weigant J et al.,2025	parallel	Germany	Healthy adults	Normotensive	≥18	4 weeks	15	15	Capsules, 200 mg cocoa flavanols/day	Cacao flavanols-free capsules (MX)	SBP, DBP, PWV, ABI	Reported (no AE)
Gröne M et al.,2023	parallel	Germany	Healthy adults	Hypertensive	55−79	30 days	35	33	Capsules, 1,000 mg cocoa flavanols / day ((–)-epicatechin 160 mg/day)	Cocoa flavanol-free capsules	SBP, DBP, FMD	NR
Siransy-Balayssac E et al., 2021	parallel	Ivory Coast	Healthy adults	Normotensive	18−30	3 weeks	25	24	Powder, 1,680 mg cocoa flavanols / day	No-intervention control	SBP, DBP	Reported (details not specified)
Garcia-Yu IA et al., 2020	parallel	Spain	Healthy postmenopausal women	Normotensive	50−64	6 months	71	66	Chocolate, 65.4 mg cocoa polyphenols/day	No-intervention control	SBP, DBP, PWV, ABI	Reported (AE)
Nishiwaki M et al., 2019	parallel	Japan	Healthy adults	Normotensive	20.7 ± 0.3	4 weeks	16	16 (cfPWV Analysis: 15)	Chocolate, 508 mg cocoa polyphenols/day	No-intervention control	SBP, DBP, PWV	Reported (no AE)
Rodriguez-Mateos A et al., 2018	parallel	Germany	Healthy adults	Normotensive	18−35	1 month	DP1-10 (High): 15 DP2-10 (Low): 15	15	Capsules with different doses of (-)-epicatechin and procyanidins High-dose group: (–)-epicatechin 130 mg; procyanidins 560 mg/day Low-dose group: (–)-epicatechin 20 mg; procyanidins 540 mg/day	Cocoa flavanol-free capsules	SBP, DBP, FMD, PWV	Reported (no AE)
Dicks L et al., 2018	parallel	Germany	Adults with type 2 diabetes and hypertension	Hypertensive	64.2 ± 1.5	12 weeks	17	18	Capsules, 207.5 mg cocoa flavanols/day	Cocoa flavanol-free capsules	SBP, DBP	Reported (no AE)
De Palma R et al., 2016	crossover	United Kingdom	Adults with chronic heart failure	Normotensive	70 ± 10	4 weeks	24	24	Chocolate, 1,064 mg cocoa flavanols/day	Chocolate, 88 mg cocoa flavanols/day	SBP, DBP	Reported (AE)
Sansone R et al., 2015	parallel	Germany	Healthy adults	Normotensive	35−60	1 month	50	50	Beverage, 900 mg cocoa flavanols/day	Cocoa flavanol-free beverage	SBP, DBP, FMD, PWV	Reported (no AE)
Rostami A et al., 2015	parallel	Iran	Adults with type 2 diabetes and hypertension	Hypertensive	35−70	8 weeks	32	28	Chocolate, 450 mg cocoa flavanols/day	Cocoa flavanol-free chocolate	SBP, DBP	NR
Massee LA et al., 2015	parallel	Australia	Healthy adults	Normotensive	18−40	4 weeks	19	19	Tablets, 250 mg cocoa flavanols/day	Cocoa flavanol-free tablets	SBP, DBP	NR
Heiss C et al., 2015	parallel	Germany	Healthy adults (Old: including cardiovascular risk factors)	Normotensive	Young (≤35): 26 ± 1 Old (50−80): 60 ± 2	14 days	Young: 11 Old: 10	Young: 11 Old: 10	Beverage, 900 mg cocoa flavanols/day	Cocoa flavanol-free beverage	SBP, DBP, FMD, PWV	Reported (no AE)
Rull G et al., 2015	crossover	United Kingdom	Adults with mild–moderate hypertension	Hypertensive	45−70	6 weeks	21	21	Chocolate, 1,064 mg cocoa flavanols/day	Chocolate, 88 mg cocoa flavanols/day	SBP, DBP	NR
Mastroiacovo D et al., 2015	parallel	Italy	Older adults with stable cardiometabolic disease	Hypertensive	61−85	8 weeks	HF (High): 30 IF (intermediate): 30	30	Beverages with different doses of cocoa flavanols High-dose group: cocoa flavanols 993 mg/day Intermediate-dose group: cocoa flavanols 520 mg/day	Beverage, 48 mg cocoa flavanols/day	SBP, DBP	Reported (AE)
Munguia L et al., 2015	parallel	Mexico	Overweight adults with multiple metabolic syndrome risk factors	Normotensive	20−60	4 weeks	10	5	Powder, 80 mg cocoa flavonoid/day	Cocoa flavonoid-free powder	SBP, DBP	Reported (no AE)
Almoosawi S et al., 2012	crossover	United Kingdom	Healthy adults	Normotensive	NR	4 weeks	BMI<25: 21 BMI≧25: 21	BMI<25: 21 BMI≧25: 21	Chocolate, 500 mg cocoa polyphenols/day (epicatechin and catechin 18.99 mg/day)	Cocoa polyphenol-free chocolate	SBP, DBP	NR
Tzounis X et al., 2011	crossover	United Kingdom	Healthy adults	Normotensive	30.2 ± 11.8	4 weeks	20	20	Beverage, 494 mg cocoa flavanols/day	Beverage, 29 mg cocoa flavanols/day	SBP, DBP	Reported (no AE)
Davison K et al., 2010	parallel	Australia	Adults with untreated mild–borderline hypertension	Hypertensive	Control: 53.0 ± 6.7 Low: 56.2 ± 14.2 Middle: 60.2 ± 13.7 High: 56.8 ± 9.7	6 weeks	High: 13 Middle: 13 Low: 12	14	Beverages with different doses of cocoa flavanols Low-dose group: cocoa flavanols 372 mg/day Middle-dose group: cocoa flavanols 712 mg/day High-dose group: cocoa flavanols 1,052 mg/day	Beverage, 33 mg cocoa flavanols/day	SBP, DBP	Reported (AE)
Muniyappa R et al., 2008	crossover	United States	Adults with essential hypertension	Hypertensive	21−65	2 weeks	20	20	Beverage, 900 mg cocoa flavanols/day	Beverage, 28 mg cocoa flavanols/day	SBP, DBP	Reported (AE)
Davison K et al., 2008	parallel	Australia	Overweight/obese adults (exercise-free arms only)	Normotensive	18−65	12 weeks	12	11	Beverage, 902 mg cocoa flavanols/day	Beverage, 36 mg cocoa flavanols/day	SBP, DBP, FMD	NR
Crews WD Jr et al., 2008	parallel	United States	Healthy older adults	Normotensive	60 or more	6 weeks	45	43	Chocolate and beverage, 754.71 mg total proanthocyanidins/day	Chocolate and beverage, 41.07 mg total proanthocyanidins/day	SBP, DBP	Reported (AE)
Al-Faris NA et al., 2008	parallel	Saudi Arabia	Healthy adults	Normotensive	18−25	15 days	30	30	Chocolate, 10.62 mg cocoa flavanols/day	Cocoa flavanol-free chocolate	SBP, DBP	NR
Taubert D et al., 2007	parallel	Germany	Adults with prehypertension or mild hypertension	Hypertensive	56−73	18 weeks	22	22	Chocolate, 28.3 mg cocoa flavanols/day	Cocoa flavanol-free chocolate	SBP, DBP	Reported (no AE)
Wang-Polagruto JF et al., 2006	parallel	United States	Postmenopausal women with hypercholesterolemia	Hypertensive	High CF: 57.7 ± 2.2 Low CF: 55.4 ± 1.7	6 weeks	9	8	Beverage, 446 mg cocoa flavanols/day	Beverage, 43 mg cocoa flavanols/day	SBP, DBP, FMD	NR
Grassi D et al., 2005	crossover	Italy	Healthy adults	Normotensive	33.9 ± 7.6	15 days	15	15	Chocolate, 500 mg cocoa polyphenols/day	Cocoa polyphenol-free chocolate	SBP	NR

1) Abbreviations: ABI, ankle–brachial index; AIx, augmentation index; BP, blood pressure; cfPWV, carotid–femoral pulse wave velocity; baPWV, brachial–ankle pulse wave velocity; DBP, diastolic blood pressure; FMD, flow-mediated dilation; PWV, pulse wave velocity; SBP, systolic blood pressure; AE, adverse event(s); NR, not reported; MX, methylxanthines.

2) Population/health status wording was standardized for clarity.

3) Baseline BP category was classified using baseline BP values and study descriptions (arm-level baseline values when reported by arm).

4) Flavanol/polyphenol amounts reflect the metrics reported in the original publications and were not converted across metrics.

5) “Outcomes (review-relevant)” indicates outcomes included in the quantitative synthesis (SBP, DBP, FMD, PWV, ABI); outcomes measured but not analyzable for this review are not listed.

6) PWV includes cfPWV, baPWV, or other PWV measures depending on the study (see Supplementary Table S2 for details).

7) For crossover trials, the same participants received both interventions.

8) Safety assessment: NR, not reported; AE, adverse events reported; no AE, explicitly stated as “no adverse events”; details not specified, safety monitoring described but AE details not reported.

Adverse events were reported in several studies; however, the reporting methods and level of detail varied substantially. In many studies, no adverse events related to cocoa flavanol intake were observed. In contrast, some studies reported mild to moderate adverse events, including gastrointestinal symptoms, as well as withdrawals due to adverse events; however, a causal relationship with the intervention was not always clearly established. In addition, safety information or adverse events were not reported in several studies.

3.3. Risk of bias within studies

The risk of bias was assessed for each outcome using RoB 2 (parallel-group trials) and the RoB 2 extension for crossover trials, and for each comparison when there were multiple comparisons within the same study. For SBP, overall risk of bias was evaluated as low risk in 5 comparisons, some concerns in 19 comparisons, and high risk in 7 comparisons. For DBP, overall risk of bias was evaluated as low risk in 5 comparisons, some concerns in 18 comparisons, and high risk in 7 comparisons. The secondary outcomes were low risk (three comparisons), some concerns (four comparisons), and high risk (one comparison) for FMD; low risk (four comparisons) and some concerns (four comparisons) for PWV; and low risk (one comparison) and some concerns (one comparison) for ABI. Adjudication details are shown in Supplementary Figure S1-S3.

3.4. Effects of cocoa flavanols on blood pressure (Primary outcomes)

Effect estimates and pooled effects for each study are shown in the forest plot (Fig. 2). For SBP and DBP, individual data for each study used in the meta-analysis (number of participants, baseline values, follow-up values, changes, and measures of variance) are shown in Supplementary Tables S4–S5.

Fig. 2.

Forest plots of primary and secondary outcomes.

Forest plots showing pooled effects of cocoa flavanols compared with control for (A) systolic blood pressure (SBP), (B) diastolic blood pressure (DBP), (C) flow-mediated dilation (FMD), (D) pulse wave velocity (PWV), and (E) ankle–brachial index (ABI). Pooled effects are presented as mean differences (MDs) with 95% confidence intervals (CIs) using a random-effects model.

SBP was pooled from 25 reports (n = 1,167). A random-effects model showed a significant reduction in SBP (MD = −2.734 mmHg, 95% CI [−3.446, −2.023], p < 0.001; Fig. 2A), with no evidence of heterogeneity (Q = 24.30, p = 0.758; I² = 0%, τ² = 0).

DBP was pooled from 24 reports (n = 1,152). DBP was significantly reduced (MD = −1.848 mmHg, 95% CI [−2.725, −0.971], p < 0.001; Fig. 2B), with evidence of moderate heterogeneity (Q = 44.64, p = 0.032; I² = 35.03%, τ² = 1.5).

3.5. Effects on vascular function indices (Secondary outcomes)

Individual data for FMD, PWV, and ABI are presented in Supplementary Tables S6–S8. FMD increased significantly when six reports (n = 324) were pooled (MD = 1.314%, 95% CI [1.043, 1.585], p < 0.001; Fig. 2C), with little evidence of heterogeneity (Q = 7.89, p = 0.342; I² = 11.28%, τ² = 0.019). PWV decreased significantly when six reports (n = 386) were pooled (MD = −0.200 m/s, 95% CI [−0.393, −0.007], p = 0.042; Fig. 2D), with no evidence of heterogeneity (Q = 2.68, p = 0.913; I² = 0%, τ² = 0). A sensitivity analysis excluding the study contributing only baPWV data yielded similar results, with the pooled effect remaining statistically significant. ABI was pooled from two reports (n = 167) and did not change significantly (MD = 0.003, 95% CI [−0.021, 0.028], p = 0.793; Fig. 2E), with no evidence of heterogeneity (Q = 1.15, p = 0.284; I² = 12.94%, τ² = 0).

3.6. Subgroup analyses

For the primary outcome, subgroup analyses were performed based on the presence of hypertension and the other health status. For SBP, pooled MD was −3.089 mmHg (95% CI [−3.941, −2.238]) in participants with hypertension (9 studies), −1.928 mmHg (95% CI [−3.282, −0.575]) in normotensive participants with or without other diseases (16 studies), and −2.313 mmHg (95% CI [−3.837, −0.789]) in normotensive healthy participants (13 studies), in the same direction as the main analysis (Fig. 3A-C).

Fig. 3.

Subgroup analyses of blood pressure.

Forest plots of subgroup analyses for systolic blood pressure (SBP) and diastolic blood pressure (DBP): (A–C) SBP and (D–F) DBP. Subgroups were defined as (A, D) participants with hypertension, (B, E) normotensive participants with or without other diseases, and (C, F) normotensive healthy participants. Effects are presented as mean differences (MDs) with 95% confidence intervals (CIs) using a random-effects model (inverse-variance method).

Similarly, for DBP, pooled MD was −2.256 mmHg (95% CI [−3.459, −1.053]) in participants with hypertension (9 studies), −1.601 mmHg (95% CI [−2.764, −0.438]) in normotensive participants with or without diseases (15 studies), and −1.616 mmHg (95% CI [−2.709, −0.523]) in normotensive healthy participants (12 studies), in the same direction as the main analysis (Fig. 3D-F).

3.7. Dose-response analysis

Dose–response analyses were conducted in 23 trials after excluding two trials (Nishiwaki 2019 and Grassi 2005) [32,51] that reported only total polyphenol content without a quantifiable dose of cocoa flavanols. Of these, 9 trials were classified as hypertensive populations and 14 trials as normotensive populations.

Because most included trials employed two-arm (single-dose) designs in this study, non-linear dose–response relationships could not be estimated and we explored the dose-response using a linear model. Dose–response analyses using linear models revealed a significant inverse association between SBP or DBP and flavanol intake (p = 0.0018 or p < 0.001, respectively; Supplementary Figure S4A and S4B), but did not identify clear dose thresholds within the examined intake range. In addition, a similar negative correlation was observed in both hypertensive and normotensive groups, and the differences in dose–response slopes between two groups were not significant (SBP: Z = −1.81, p = 0.070, DBP: Z = −0.85, p = 0.397).

3.8. Publication bias

Publication bias was assessed using funnel plots and Egger’s test (significance level p < 0.1). SBP showed no clear asymmetry (Fig. 4A), and the Egger’s test results were not significant (p = 0.35). DBP showed mild asymmetry (Fig. 4B), and Egger’s test suggested a significant difference (p = 0.08). The pooled effect was also significant when eight imputed studies were reanalyzed using the trim-and-fill method (adjusted MD = −1.12, 95% CI [−2.05, −0.193]; Supplementary Figure S5). The Egger’s test was not significant for FMD (p = 0.38) or PWV (p = 0.27). Egger’s test could not be performed for ABI because the number of studies was small and the included studies had standard errors of 0. Outcomes with <10 pooled studies were interpreted cautiously due to limitations in the reliability of these assessments.

Fig. 4.

Funnel plots for publication bias assessment.

Funnel plots for (A) systolic blood pressure (SBP) and (B) diastolic blood pressure (DBP) were used to assess small-study effects and publication bias.

3.9. Certainty of evidence (GRADE)

The GRADE-based certainty assessment was moderate for SBP and low for DBP. Secondary outcomes were low for FMD and PWV and very low for ABI. A summary of the findings (SoF) and downgrade rationale is provided in Supplementary Tables S9-S10.

4. Discussion

In this systematic review and meta-analysis, cocoa flavanol intake significantly reduced SBP and DBP. In addition, cocoa flavanol intake was associated with improvements in endothelial function (FMD) and arterial stiffness (PWV), providing an integrated assessment of both functional and mechanical vascular effects. Because elevated blood pressure is linked to CVD risk [52], and endothelial dysfunction and arterial stiffness are associated with future cardiovascular events [5,9], these findings support the potential role of cocoa flavanols in cardiovascular health.

The strength of this study is the comprehensive evaluation of cocoa flavanols on both blood pressure and multiple indices of vascular function, including FMD, PWV, and ABI, within a single quantitative synthesis. By integrating these outcomes, the present study provides a broader assessment of vascular health that extends beyond blood pressure alone and offers broader insight into vascular function relevant to cardiovascular disease risk. Our findings are generally consistent with previous systematic reviews and meta-analyses reporting modest blood pressure-lowering effects of cocoa flavanols [[18], [19], [20]]. In this study, significant reductions in both SBP and DBP were also observed with cocoa flavanol intake. However, previous meta-analyses primarily focused only on blood pressure or cardiometabolic outcomes, whereas comprehensive evaluations integrating multiple vascular function indices have been limited [[18], [19], [20]]. With the incorporation of the latest randomized controlled trials, this study enabled a broader assessment and a more comprehensive evaluation of both functional and mechanical vascular effects. The expanded data also enabled subgroup analyses according to blood pressure status and exploratory dose–response analyses, providing further insight into the potential cardiovascular effects of cocoa flavanols.

Furthermore, the blood pressure–lowering effect of cocoa flavanols was observed across all subgroups (participants with hypertension, normotensive participants with or without other diseases, and normotensive healthy participants), consistent with the main analysis. Therefore, the robustness of the primary outcome results was supported by the fact that the direction and magnitude of the effect estimates did not differ between the subgroup analyses and the main analysis. While there were larger point estimates in hypertensive participants than normotensive participants, differences between subgroups were not formally tested and should be interpreted cautiously. In addition, between-study heterogeneity was low for SBP, FMD, and PWV and moderate for DBP, suggesting generally consistent findings across studies. These findings also support the robustness of the pooled effects on the primary outcomes. Accordingly, cocoa flavanols may be considered as a potentially complementary dietary approach across adult populations, including those at higher cardiovascular risk.

The findings in this study are consistent with previously reported mechanisms by which cocoa flavanols lower blood pressure. Cocoa flavanols are thought to lower blood pressure by dilating blood vessels and reducing peripheral vascular resistance through endothelial NO production, and the improvement in FMD reflects this mechanism [53,54]. The FMD benefit of epicatechin- and flavanol-rich cocoa was lost with pretreatment using NG-nitro-L-arginine methyl ester, supporting a nitric oxide synthase-dependent mechanism [15,53,54]. In addition, the observed effects were generally consistent in direction across studies. These findings indicate that the blood pressure–lowering and vascular effects could be expected within the range examined.

We have also observed the significant inverse association between flavanol intake and blood pressure, but this does not allow conclusions about dose thresholds or minimal effective intakes. This is because non-linear dose–response patterns could not be examined, which may be physiologically plausible. Furthermore, a recent dose–response meta-analysis of cocoa consumption reported that significant effects on anthropometric outcomes were primarily observed in interventions lasting more than 4 weeks [55]. However, in the present study, exploratory subgroup analyses using the same 4-week cut-off showed significant reductions in both duration categories, and no statistically significant subgroup differences were observed (SBP: p = 0.079; DBP: p = 0.425). Further research is needed, especially with multiple dose levels in order to analyze nonlinear dose-response patterns.

The reductions of approximately 2.7 mmHg in SBP and 1.8 mmHg in DBP by cocoa flavanols appear modest. However, epidemiological and clinical evidence suggests that blood pressure is continuously associated with cardiovascular risk, including stroke, without any threshold, and that even small reductions in mean blood pressure may reduce the risk burden in the long term [52,56]. For example, an approximately 3 mmHg lower SBP has been associated with reductions in stroke mortality (8%), coronary heart disease mortality (5%), and overall mortality (4%) at the population level [57]. Similarly, a 2 mmHg lower DBP has been associated with lower risks of coronary heart disease and stroke [58]. These epidemiological observations provide context for the potential public health relevance of modest blood pressure reductions, although it should be noted that they do not represent direct evidence of the effects of cocoa flavanol intake on cardiovascular events or mortality. Large meta-analyses of RCTs have consistently shown that greater blood pressure reductions are associated with greater reductions in major cardiovascular events [56]. In older adults, impaired endothelial function and increased arterial stiffness have also been reported to be associated with frailty and sarcopenia [59]. Therefore, the blood pressure-lowering effect of an implementable dietary intervention, such as cocoa flavanol intake, may have public health implications [58].

There was no significant effect for ABI in this study. Interpretation is limited because ABI was available from only two studies and the certainty of evidence was very low, resulting in substantial imprecision. Longer follow-up may be required to detect effects on ABI, which may reflect more slowly changing vascular properties than blood pressure, FMD, or PWV. Besides, future trials should include longer follow-up, prespecified dose/formulation conditions, and standardized vascular assessment protocols to improve interpretability across studies. Where feasible, incorporating ABPM may further strengthen measurement reliability and clinical relevance [60], in addition to ongoing efforts to standardize vascular assessment such as FMD and PWV [6].

This study has three limitations. First, the certainty of evidence (GRADE) was moderate for SBP but low for DBP, FMD, and PWV and very low for ABI, mainly due to imprecision and the limited number of available studies for several outcomes. Accordingly, findings for vascular indices—particularly ABI—should be interpreted cautiously. Second, dietary intervention RCTs are susceptible to bias due to challenges in placebo control and blinding, crossover carryover, control for concurrent interventions (e.g., antihypertensive drugs, salt intake, and weight loss), and standardization of measurements (e.g., differences in FMD procedures and PWV instruments). In addition, residual confounding, such as salt intake and antihypertensive drug use, cannot be fully controlled. All of these may affect the interpretation of BP-related outcomes. Third, dose-response and formulation differences due to flavanol content, intake form, processing degree, and food matrix (co-occurring lipids and carbohydrates) have not been fully established, which may affect the interpretation and implementation of the results. Although exploratory dose–response analyses were conducted in this review, dose, formulation, and matrix optimization should be prespecified in future studies to better characterize potential dose–response relationships and minimally effective doses. In addition, subgroup analyses according to participant age were not performed, and the influence of intervention duration was evaluated only in an exploratory manner. Because several studies included broad and overlapping age ranges, and there is no established threshold for defining short- versus long-term cocoa flavanol interventions, such analyses would have required arbitrary categorization and may have been difficult to interpret. Although adverse event reporting varied substantially across the included studies, this review was not designed to systematically evaluate safety outcomes. Therefore, no quantitative synthesis of adverse events was performed, and the safety profile of cocoa flavanol intake should be interpreted cautiously based on the available descriptive information.

5. Conclusions

This systematic review and meta-analysis showed that continuous cocoa flavanol intake significantly reduced SBP and DBP, and improved vascular function, as indicated by increased FMD and reduced PWV, within the range of doses examined in randomized controlled trials (10.6–1,680 mg). Evidence regarding ABI remains insufficient. Overall, these findings suggest that cocoa flavanols could lower blood pressure and improve endothelial function as an adjunctive nutritional intervention that is easy to implement in daily life, potentially helping to reduce CVD risk. Further well-designed trials with longer follow-up and standardized vascular assessments are warranted.

Author contributions

Conceptualization, MS, MN and KN; Data curation, NW and LH; Formal analysis, NW and LH; Funding acquisition, KN; investigation, NW and LH; Methodology, MS, NW and KN; Project administration, KN; Resources, NW; writing-original draft preparation, MS, KI and YA; writing-review and editing, MN and KN; and visualization, MS, NW, and LH. All authors have reviewed and consented to the published version of the manuscript.

Informed consent

Not applicable.

Ethical statement

Not applicable. This study does not involve human participants, human data, or human tissue.

Studies on Human and/or Animal statement

This article does not report any new studies involving human participants or animals performed by the authors. All data analyzed in this study were obtained from previously published studies.

Declaration of Generative AI and AI-assisted technologies in the writing process

During the preparation of this work, the authors used OpenAI’s ChatGPT in order to improve the language of this manuscript. After using this tool/service, the authors reviewed and edited the content as needed and take full responsibility for the content of the published article.

Funding

This research received funding from Meiji Co., Ltd.

Approval and registration number

Ethical approval was not required because this systematic review and meta-analysis was based exclusively on data from previously published studies. The review protocol was prospectively registered with the University Hospital Medical Information Network Clinical Trials Registry (UMIN-CTR; Registration No. UMIN000059995).

Data statement

The data supporting the findings of this systematic review and meta-analysis are available within the article and its Supplementary Materials. Additional information is available from the corresponding author upon reasonable request.

Declaration of competing interest

M.S., K.I., Y.A., M.N., and K.N. are employees of Meiji Co., Ltd. L.H. and N.W. are employees of Yakujihou Marketing Jimusho Co., Ltd. Meiji Co., Ltd. is involved in the research and development of food products, including cocoa-containing foods. However, neither Meiji Co., Ltd. nor Yakujihou Marketing Jimusho Co., Ltd. had any role in the production, marketing, or distribution of specific cocoa flavanol–containing products evaluated in this study. The analyses and interpretations presented herein were conducted independently.

Appendix A. Supplementary data

The following is Supplementary data to this article: