Coffee Consumption and Risk of Hypertension in Adults: Systematic Review and Meta-Analysis

Objectives: The association between coffee intake and hypertension (HTN) risk is controversial. Therefore, this systematic review and meta-analysis aimed at summarizing the current evidence on the association of coffee with hypertension risk in observational studies. Methods: PubMed/Medline and Web of Science were searched for observational studies up to February 2023. Observational studies which assessed the risk of HTN in the highest category of coffee consumption in comparison with the lowest intake were included in the current meta-analysis (registration number: CRD42022371494). The pooled effect of coffee on HTN was evaluated using a random-effects model. Results: Twenty-five studies i.e., thirteen cross-sectional studies and twelve cohorts were identified to be eligible. Combining 13 extracted effect sizes from cohort studies showed that higher coffee consumption was associated with 7% reduction in the risk of HTN (95% CI: 0.88, 0.97; I2: 22.3%), whereas combining 16 effect sizes from cross-sectional studies illustrated a greater reduction in HTN risk (RR = 0.79, 95% CI: 0.72, 0.87; I2 = 63.2%). These results varied by studies characteristics, such as the region of study, participants’ sex, study quality, and sample size. Conclusions: An inverse association was found between coffee consumption and hypertension risk in both cross-sectional and cohort studies. However, this association was dependent on studies characteristics. Further studies considering such factors are required to confirm the results of this study.


Introduction
Hypertension is directly associated with the risk of cardiovascular diseases [1]. The number of people aged 30 to 79 years with hypertension has increased from 648 million in 1990 to more than 1.2 billion people in 2019 [2], making it a serious public health concern, especially in low and middle-income countries [3]. Additionally, evidence suggests that high-quality diets are responsible for a 22% decline in the risk of cardiovascular disease [3,4], while the consumption of red and processed meats, high sodium intake, low potassium intake, obesity, alcohol consumption, as well as sugar-sweetened beverages are associated with an increase in the risk of hypertension [1,5].

Methods
Search Strategy and Study Selection: This meta-analysis was designed, analyzed, and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement. A systematic search of the available publications was performed using the MEDLINE and Web of Science from inception (1952) to February 2023. Without any restriction on publication date and language, a search strategy applying the key terms of "(coffee OR caffeine) AND (hypertension OR "blood pressure")" was conducted. No restriction was placed on the article language and publication date. A manual search of recent reviews and relevant original articles was performed for additional relevant studies. The study protocol was registered on the International Prospective Register of Systematic Reviews, PROSPERO (registration number: CRD42022371494).
Study selection: Two independent reviewers (FH and PH) screened titles and abstracts for eligible articles. Disagreements were resolved by discussion with C.d.O. and E.S. Then, based on the full text of identified articles, their eligibility for inclusion was assessed according to our inclusion and exclusion criteria. Studies were included if they met the following criteria: (1) original studies on adult population (aged 18 or older), (2) an observational design (prospective cohort, case-control, or cross-sectional), (3) reporting the relative risk (RR) or hazard ratio (HR) or odds ratio (OR) with 95% confidence interval (CI), and (4) assessment of hypertension or elevated blood pressure in subjects with the highest vs. the lowest intake of coffee, irrespective of coffee types and their caffeine content. Studies were excluded if they were not original research or were in vitro or animal model, conducted on adolescents or children, examined gestational or ambulatory or coffee post injection blood pressure or hypertension risk, reported blood pressure mean, assessed blood pressure control over the time, and examined HTN risk for coffee polyphenols. Disagreements were resolved through discussion until agreement was reached.
Data extraction and quality assessment: Using a pre-designed extraction form, two independent reviewers (PH and AM) extracted the following information: the author's first name, country, year, study design, mean age or age range of participants, sample size, follow-up duration for longitudinal studies, instruments used to assess coffee intakes, the method and cut-off point used to HTN diagnosis, main findings (effect and 95% CI), and adjusted confounders. If the results of studies were reported in various groups, the information of all groups was extracted. The quality of included studies in our meta-analysis was assessed using the New Castel Ottawa Scale (NOS) designed for observational studies. In general, this scale consisted of 3 main domains (selection, comparability, and outcome) and 8 questions in total for cohort studies. The minimum score for each domain is zero, while the maximum score varies between domains. The highest scores for participants' selection, participants' comparability, and assessment of outcome/exposure domains are 4, 2, and 3, respectively. The overall score ranges from zero to nine. In the adapted version for cross-sectional studies, consisting of 7 questions, the maximum score is 10. In the present study, the quality scores of ≥7 was considered methodologically high, and those with the score of 6 or fewer were considered low methodological quality [17,18].

Statistical Analysis
Hypertension risk was reported either as relative risk (RR) or odds ratio (OR). Due to the high prevalence of HTN (32% in women and 34% in men) [2], ORs were converted into RRs [19]. To combine effect sizes, a random-effects model on the basis inverse-variance method, which incorporate between-study heterogeneity [20,21] was used. Potential between-studies heterogeneity was estimated using I 2 values. Heterogeneity was considered substantial when I 2 values were greater than 50% [21]. Heterogeneity sources were examined using sub-group analysis based on participants' sex, study design, geographical region of studies, participants age (< vs. >50 years), sample size (< vs. >8000 in cohort studies and < vs. >3000 in cross-sectional studies), hypertension stage (stage I and II: SBP ≥ 130 mmHg and/or DBP ≥ 80 or 85 mmHg, and only stage II: SBP ≥ 140 mmHg and/or DBP ≥ 90 mmHg) [22], study quality score (≤7 vs. >7) [23], and follow-up duration for cohort studies.
Publication bias was tested using visual inspection of a funnel plot, Egger's test, and Begg's tests [24,25]. Sensitivity analysis was performed to determine to which extent each individual study influences the pooled effect. All analyses were performed using the Stata 11.0 software. p values < 0.05 were regarded as significant.

Search Results
The flowchart of study selection process is illustrated in Figure 1. Our search strategy identified 2039 articles. After removal of duplicate articles (n = 128), 1911 articles were screened for their titles and abstracts, of which 84 articles remained for further evaluation based on their full texts. Finally, 25 articles, published between 2002 and 2023, met our inclusion criteria and were included in the present meta-analysis . Four of the twenty-five eligible articles were considered two separate studies or populations since they either consisted of two separate cohorts [45], or reported the results separately for men and women [48,49] or based on genes variants [40].

Cohort Studies
The pooled results of the meta-analysis revealed that individuals with the highest coffee consumption in comparison with the lowest intake had 7% lower risk for HTN (RR = 0.93, 95% CI: 0.88, 0.97; I 2 : 22.3%) (Figure 2). All selected cohort studies evaluated coffee consumption through a specific questionnaire. Findings from the sensitivity analysis demonstrated no change in the significance of the findings by excluding an individual study at the time. Despite a small asymmetry in the funnel plot, the Egger test (p = 0.449) and Begg test (p = 0.714) revealed no evidence of publication bias (Supplementary Figure S1).

Cohort Studies
The pooled results of the meta-analysis revealed that individuals with the highest coffee consumption in comparison with the lowest intake had 7% lower risk for HTN (RR = 0.93, 95% CI: 0.88, 0.97; I 2 : 22.3%) (Figure 2). All selected cohort studies evaluated coffee consumption through a specific questionnaire. Findings from the sensitivity analysis demonstrated no change in the significance of the findings by excluding an individual study at the time. Despite a small asymmetry in the funnel plot, the Egger test (p = 0.449) and Begg test (p = 0.714) revealed no evidence of publication bias (Supplementary Figure  S1). When studies were stratified based on the geographical region of the study, higher consumption of coffee compared with the lowest was associated with lower risk of HTN in studies which were conducted in the United States (RR = 0.92, 95% CI: 0.87, 0.97; I 2 = 0.7%). However, it was not associated in Europe (RR = 0.97, 95% CI: 0.83, 1.13; I 2 = 60.5%) and Asia (RR = 0.94, 95% CI: 0.83, 1.07; I 2 = 25.0%). The results of subgroup analysis demonstrated an inverse association in the subgroups of studies with a follow-up duration of at least 10 years (RR = 0.91, 95% CI: 0.84, 0.99), sample size less than median (RR = 0.89, 95% CI: 0.81, 0.99) and cases more than median (RR = 0.93, 95% CI: 0.89, 0.97), and those which were low quality (RR = 0.88, 95% CI: 0.83, 0.94) or conducted in females only (RR = 0.93, 95% CI: 0.88, 0.98), while in the counterpart groups a null association was found. Subgroup analysis based on age showed similar results in both groups (Table 3). When studies were stratified based on the geographical region of the study, higher consumption of coffee compared with the lowest was associated with lower risk of HTN in studies which were conducted in the United States (RR = 0.92, 95% CI: 0.87, 0.97; I 2 = 0.7%). However, it was not associated in Europe (RR = 0.97, 95% CI: 0.83, 1.13; I 2 = 60.5%) and Asia (RR = 0.94, 95% CI: 0.83, 1.07; I 2 = 25.0%). The results of subgroup analysis demonstrated an inverse association in the subgroups of studies with a follow-up duration of at least 10 years (RR = 0.91, 95% CI: 0.84, 0.99), sample size less than median (RR = 0.89, 95% CI: 0.81, 0.99) and cases more than median (RR = 0.93, 95% CI: 0.89, 0.97), and those which were low quality (RR = 0.88, 95% CI: 0.83, 0.94) or conducted in females only (RR = 0.93, 95% CI: 0.88, 0.98), while in the counterpart groups a null association was found. Subgroup analysis based on age showed similar results in both groups (Table 3).

Cross-Sectional Studies
In cross-sectional studies, compared with the lowest amount of coffee consumption, the highest amount was associated with lower risk of HTN (OR = 0.79, 95% CI: 0.72, 0.87). Heterogeneity was considerably high between studies (I 2 : 63.2%) (Figure 3). Findings from the sensitivity analysis demonstrated no change in the significance of the findings by excluding an individual study at the time. No evidence of publication bias was observed (p for the Egger test = 0.198 and for the Begg test = 0.753).
In cross-sectional studies, compared with the lowest amount of coffee consumption, the highest amount was associated with lower risk of HTN (OR = 0.79, 95% CI: 0.72, 0.87). Heterogeneity was considerably high between studies (I 2 : 63.2%) (Figure 3). Findings from the sensitivity analysis demonstrated no change in the significance of the findings by excluding an individual study at the time. No evidence of publication bias was observed (p for the Egger test = 0.198 and for the Begg test = 0.753).  Table 3).

Discussion
The results of the present meta-analysis suggest, overall, a slight reduction in the risk of HTN following coffee consumption. This association was observed in both cohort and cross-sectional studies, and was much stronger in cross-sectional studies. However, this association was influenced by various factors in cohort and cross-sectional studies, such as geographical region, sex, sample size, and study quality.
There are several meta-analyses evaluating the relationship between coffee and HTN risk [12,51,52]. In the most recent meta-analysis, published in November 2022, 12 prospective cohort studies were included, and a null association was found [53]. However, this study has some methodological limitations. First, they included two studies which were conducted on hypertensive individuals. Second, they did not convert OR into RR in two studies despite the high prevalence of HTN. Third, the authors failed to include two relevant studies. In addition, another earlier meta-analysis on four prospective cohort studies  Table 3).

Discussion
The results of the present meta-analysis suggest, overall, a slight reduction in the risk of HTN following coffee consumption. This association was observed in both cohort and cross-sectional studies, and was much stronger in cross-sectional studies. However, this association was influenced by various factors in cohort and cross-sectional studies, such as geographical region, sex, sample size, and study quality.
There are several meta-analyses evaluating the relationship between coffee and HTN risk [12,51,52]. In the most recent meta-analysis, published in November 2022, 12 prospective cohort studies were included, and a null association was found [53]. However, this study has some methodological limitations. First, they included two studies which were conducted on hypertensive individuals. Second, they did not convert OR into RR in two studies despite the high prevalence of HTN. Third, the authors failed to include two relevant studies. In addition, another earlier meta-analysis on four prospective cohort studies conducted on general population [12] revealed a null association for habitual coffee consumption and HTN. However, we identified 6 further studies published after 2017 which were not included in their meta-analysis [27,[30][31][32]36,47].
Overall, this meta-analysis suggested an inverse association between coffee consumption and HTN risk in both cross-sectional and cohort studies with a small heterogeneity in prospective studies. There is a large discrepancy between studies with regard to the association of coffee and HTN risk. While some studies suggested a favorable effect [26,28,29,31], others have reported a null or an adverse association [27,37,39,40]. The plausible mechanisms underpinning the inverse association between coffee and HTN risk might be attributable to high levels of antihypertensive nutrients (i.e., vitamin E, niacin, potassium and magnesium) and polyphenols in coffee [54]. These factors can modulate blood pressure through their antioxidant and anti-inflammatory properties, and effect on nitric oxide synthesis, lipid metabolism, and endothelial function [55,56]. Moreover, caffeine and chlorogenic acid (CGA) in coffee play a role in insulin homeostasis [57]. Insulin, in turn, causes sodium retention which activates symptomatic nervous system activity and the proliferation of vascular smooth muscle, leading to higher levels of blood pressure [58,59]. Other anti-hypertensive effects of CGA include its anti-inflammatory property [60], inhibitory angiotensin-converting enzyme activity [61], and vasodilation effect through increasing nitric-oxide bioavailability [62].
Our subgroup analyses indicated that the association between coffee and HTN risk differs by countries. Similar to an earlier meta-analysis, we found an in inverse association only in the USA, but not in Europe and Asia in cohort studies. This difference might be explained, at least to some extent, by the amount of coffee consumed. The inverse association between coffee and HTN risk is a non-linear association, and a coffee consumption less than 3 cups/day could not decrease HTN risk [12]. Based on the available evidence, the amount of coffee consumed in European and Asian countries was considerably lower than that of in the USA in 2016 [63]. In addition, the null association in Asian and European countries might be mediated by other variables, such as genetic factors [64], the method used to prepare the coffee, the kind of consumed coffee, and smoking [51,54]. It is also possible that variations in lifestyle and dietary habits in different parts of a continent lead to different association. For instance, Hu et al. [44] failed to find any significant association between coffee and HTN risk in Finland as a North European country, whereas Navarro et al. found an inverse association in Spain as a South European country. However, due to the small number of studies containing such information or located in a specific region, we could not further explore the reason behind this finding.
The characteristics of studies, such as study quality, number of cases, and HTN definition could have affected the association. The results indicated an inverse association between coffee consumption and HTN risk in studies with a larger case number and those which used stage 1 cut-off point for HTN definition. A smaller case number cannot provide enough power to find the expected effect size, which is in concordance with our results as the confidence interval for studies with smaller case number was wider than those with more cases [65].
There was only one study which found a positive direct link between coffee consumption and HTN risk [33]. This study was conducted amongst patients with type 1 diabetes and approximately 50% of their study population consumed antihypertensive medications, which was considered a criterion for having HTN. This might restrict the generalizability of their findings per se.
The main strengths of this meta-analysis are its large population from different countries and ethnicities with a wide age range, a comprehensive literature search to identify relevant articles, and subgroup analysis based on various confounders. Our findings were robust, and no evidence of publication bias was observed. In addition, most cohort studies were long enough to allow investigating incidence of hypertensive cases. We have also investigated a wide range of characteristics of studies which can influence the coffee-HTN association. The first and the main limitation of our study which should be considered for the future studies is the lack of adequate data for determining other plausible heterogeneity sources, such as sodium intake, smoking status, and the type of coffee and its preparing method. For instance, Grosso et al. showed that the association differs by the smoking status of participants. While in the whole population, they reported a null association with a wide confidence interval, in stratified analysis by smoking, they found an inverse association. They found that both non-smoker males and females who drank 3-4 cups coffee/d had lower risk of HTN development, whereas no significant association was found in smokers [37]. Furthermore, it is plausible that in countries which mainly tend to consume instant coffee, coffee consumption exhibits more favorable effects on health status due to its greater antioxidant activities in comparison with brewed coffee [66]. Second, between-studies inconsistency in terms of the adjusted confounders is another limitation which may explain between-studies heterogeneity, and should be considered. Third, estimating the amount of coffee consumption using different approaches, with their own specific measurement errors, may cause misclassification due to the combination of various errors. This, in turn, can affect the accuracy of findings. Fourth, heterogeneity was substantially high in the cross-sectional analysis and it could not be eliminated using subgroup analysis. Therefore, our findings should be interpreted cautiously.
In conclusion, this meta-analysis showed an inverse association between coffee consumption and hypertension incidence, either cross-sectionally or prospectively. However, this association is dependent on the geographical region of the study, participants' sex, the number of cases, and study quality. Future studies with an appropriate design which consider the effect of other confounders on HTN risk are warranted to confirm this result.