Introduction

Atrial fibrillation presents a considerable public health burden and is the most common type of arrhythmia affecting around 1–2% of the general population, increasing to around 10% of persons by 80 years of age [1]. Five million incident cases were diagnosed worldwide in 2010 [2] and the prevalence of atrial fibrillation has been estimated at 33 million in 2015 [3]. In the USA the prevalence of atrial fibrillation has been projected to increase from 2.3 million in 1996–1997 to 5.6 million by 2050 [4]. Patients with atrial fibrillation are at increased risk of a number of complications, most notably stroke, heart failure, dementia and all-cause mortality [5, 6]. Several risk factors for atrial fibrillation have been established including age, sex, diabetes, coronary heart disease, heart failure, smoking, alcohol, obesity, low physical activity and possibly high intensity physical activity [7,8,9,10,11,12,13,14,15].

Elevated blood pressure is the leading cause of death and disability-adjusted life-years (DALYs) globally with 10.4 million deaths and 218 million DALYs attributable to elevated systolic blood pressure in 2017 according to the Global Burden of Disease Study [16]. Although elevated blood pressure is an established risk factor for several cardiovascular diseases, data regarding blood pressure and risk of atrial fibrillation have to our knowledge not been summarized in a meta-analysis. A large number of cohort studies have investigated the association between hypertension and the risk of atrial fibrillation [7,8,9, 17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65], and most of these found an increased risk [7, 8, 17,18,19,20,21,22, 24,25,26,27,28,29,30,31, 33, 34, 36,37,38,39,40,41,42,43,44,45,46,47,48,49, 51,52,53,54,55,56,57,58,59,60,61,62, 64, 65], with few studies reporting no association [9, 23, 32, 35, 50, 63], however, the strength of the association has differed considerably between studies with reported relative risks reported ranging from 0.93 to 2.85 [7,8,9, 17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59]. In addition, several studies have investigated the association between systolic [8, 18, 20, 29, 33, 34, 39, 41, 42, 44, 46, 51, 55, 63, 65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80] or diastolic [8, 29, 33, 39, 41, 44, 46, 51, 55, 67,68,69,70, 72,73,74, 79, 81] blood pressure and risk of atrial fibrillation with most studies reporting increased risk for increasing systolic blood pressure [18, 20, 29, 34, 39, 41, 44, 46, 51, 55, 65,66,67,68,69,70,71,72, 74,75,76,77,78,79,80], while results have been more mixed for diastolic blood pressure with some showing an increased risk [39, 44, 51, 55, 67,68,69, 72, 79, 82] but other studies showing no association [8, 29, 33, 46, 70, 73, 74, 81], or even reduced risk [41, 77] with higher diastolic blood pressure.

Establishing whether hypertension and elevated blood pressure increases the risk of atrial fibrillation would be important from a preventive point of view as it is a risk factor that could be modified by diet, physical activity, weight control and pharmaceutical drugs [83]. In addition it would be useful to better characterize the strength and shape of the dose–response relationship between blood pressure and atrial fibrillation to clarify whether the association is dose-dependent or if there are threshold effects. We conducted a systematic review and meta-analysis of cohort studies on hypertension and blood pressure in relation to the risk of atrial fibrillation to clarify the strength and shape of the dose–response relationship, and to identify potential sources of heterogeneity in the results.

Material and methods

Search strategy and inclusion criteria

We searched Pubmed, and Embase databases up to June 9th 2022 for eligible studies. The search strategy is provided in the Supplementary Text. We followed standard criteria for conducting and reporting meta-analyses [84]. In addition, we searched the reference lists of the identified publications for further studies.

Study selection

We included published retrospective and prospective cohort studies and nested case–control studies within cohorts that investigated the association between blood pressure or hypertension and the risk of atrial fibrillation (any type). Retrospective case–control studies were excluded because of their potential for recall bias and selection bias and cross-sectional studies were excluded because of difficulties in establishing cause and effect relationships. Estimates of the relative risk adjusted for at least one confounding factor had to be available with the 95% confidence intervals (CIs) in the publication. Conference abstracts, grey literature and non-English publications were not included. When multiple publications were available from the same study, the study with the largest number of cases was used in general. However, overlapping publications were used in specific subgroup analyses by sex or ethnicity, when the article used for the main analysis did not report such stratified analyses. Overlapping publications that reported risk estimates in three categories or more were also used for the nonlinear dose–response analyses (as the nonlinear analysis requires categorical data) if the article included in the main analysis only reported risk estimates on a continuous scale. A list of the excluded studies can be found in Supplementary Table 1. DA, YMS, EK and TF did the study selection in duplicate and any disagreements were resolved by discussion.

Data extraction

The following data were extracted from each study: The first author’s last name, publication year, country where the study was conducted, study period, sample size, number of cases and participants, exposure (hypertension, systolic blood pressure, or diastolic blood pressure), subgroup (e.g. sex, race), relative risks (RRs) and 95% CIs for hypertension versus no hypertension or for increments in systolic or diastolic blood pressure and variables adjusted for in the analysis. DA did the data extraction and it was checked for accuracy by YMS.

Statistical methods

We calculated summary RRs (95% CIs) of atrial fibrillation for participants with hypertension compared with participants without hypertension and for systolic and diastolic blood pressure using the random-effects model by DerSimonian and Laird [85] which takes into account both within and between study variation (heterogeneity). The average of the natural logarithm of the RRs was estimated and the RR from each study was weighted by the inverse of its variance. Linear dose–response analyses were conducted per 20 mmHg for systolic blood pressure and per 10 mmHg for diastolic blood pressure (consistent with previous studies [86,87,88]) using the method of Greenland and Longnecker [89]. For studies that reported blood pressure by ranges we estimated the midpoint for each category by calculating the average of the upper and lower cut-off points. For open-ended categories we used the width of the adjacent interval to estimate an upper or lower cut-off value for the extreme category. Fractional polynomial models were used to investigate a potential nonlinear association between systolic and diastolic blood pressure and risk of atrial fibrillation [90]. A log-likelihood test was used to test for nonlinearity [91].

Heterogeneity between studies was evaluated using Q and I2 statistics [92]. I2 is an estimate of how much of the heterogeneity that is due to between study variation rather than chance. I2-values of 25%, 50% and 75% indicates low, moderate and high heterogeneity respectively. We conducted main analyses (all studies combined) and stratified by study characteristics such as sample size, number of cases, whether prevalent cases were excluded or not, geographic location, study quality and by adjustment for confounding factors to investigate potential sources of heterogeneity and we used meta-regression analyses to test for differences in summary estimates between subgroups. Study quality was assessed using the Newcastle Ottawa scale which rates studies according to selection, comparability and outcome assessment with a score range from 0 to 9 [93].

Publication bias was assessed using Egger’s test [94] and by inspection of funnel plots. The statistical analyses were conducted using the software package Stata, version 13.1 software (StataCorp, Texas, US).

Results

From a total of 32,876 records that were identified by the search we included a total of 69 publications [7,8,9, 17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82] with data from 68 cohort studies (two of these were nested case–control studies within cohort studies [21, 66]) in the systematic review and meta-analysis of hypertension and blood pressure and atrial fibrillation (Fig. 1). Five of these publications were identified from separate searches on other risk factors for atrial fibrillation [9, 21, 32, 34, 42]. Each of two publications reported results from two studies combined [73, 81], and another publication reported results from six studies combined [65]. Two publications [42, 75] reported results from two separate studies each and one publication reported results from five separate studies [29], two of which were included in the main analysis (the other three were surpassed by more recent publications, but results of two of these duplicate studies were included in subgroup analyses by ethnicity). Twenty six studies (23 publications) [18, 20, 21, 25, 27,28,29,30, 33, 36, 42, 43, 48, 54, 64, 65, 69, 71,72,73, 75, 76, 78] were from Europe, twenty studies (25 publications) [7, 17, 19, 22, 24, 26, 29, 31, 34, 37, 40, 41, 45,46,47, 52, 53, 56, 59, 67, 68, 70, 74, 77, 81] were from North America, nineteen studies (19 publications) [23, 32, 35, 38, 39, 44, 49,50,51, 55, 57, 60,61,62,63, 66, 79, 80, 82] were from Asia, and three studies (3 publications) [8, 9, 58] were from Australia.

Fig. 1
figure 1

Flow-chart of study selection

Full size image

Fifty six cohort studies (52 publications, 52 risk estimates) [7,8,9, 17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65] were included in the analysis of hypertension and atrial fibrillation risk including 1,080,611 cases and 30,539,230 participants (Fig. 1, Table 1). Data on hypertension and atrial fibrillation by ethnicity [29, 59] and sex [25] from three studies (ARIC, REGARDS, and Malmö Diet and Cancer Study) were only included in the respective subgroup analyses as the publications overlapped with more recent publications from the same studies which were used for the main analysis [27, 45, 56]. The summary relative risk for persons with hypertension compared to persons without hypertension was 1.50 (95% CI: 1.42–1.58, I2 = 98.1%, pheterogeneity < 0.0001) (Fig. 2). There was no evidence of publication bias neither with Egger’s test (p = 0.74) or by inspection of the funnel plot (Supplementary Fig. 1). The summary RR ranged from 1.47 (95% CI: 1.42–1.52) when excluding the study by Zoller et al. [30] to 1.51 (95% CI: 1.43–1.59) when excluding the study by Sano et al. [53] (Supplementary Fig. 2).

Table 1 Cohort studies of hypertension and atrial fibrillation
Full size table
Fig. 2
figure 2

Hypertension and atrial fibrillation

Full size image

Thirty seven cohort studies (28 publications, 31 risk estimates) [8, 18, 20, 29, 33, 34, 39, 41, 42, 44, 46, 51, 55, 63, 65,66,67, 69,70,71,72,73,74,75, 77,78,79,80] (347,813 cases, 14,565,763 participants) were included in the analysis of systolic blood pressure and atrial fibrillation (Table 2). The summary RR was 1.19 (95% CI: 1.16–1.21, I2 = 68.4%, pheterogeneity < 0.0001) per 20 mmHg increment (Fig. 3a). There was no evidence of publication bias with Egger's test (p = 0.46), but some indication of asymmetry in the funnel plot (Supplementary Fig. 3a). When using the trim and fill method, eight studies were added, but the results were similar, summary RR = 1.17 (95% CI: 1.14–1.19) (Supplementary Fig. 4). The summary RR ranged from 1.18 (95% CI: 1.16–1.20) when the Atherosclerosis Risk in Communities Study [77] was excluded to 1.20 (95% CI: 1.18–1.22) when a pooled analysis [65] was excluded (Supplementary Fig. 5). Ten cohort studies [18, 34, 39, 55, 67,68,69, 76, 79, 80] were included in the nonlinear dose–response analysis. Although the test for nonlinearity was significant, pnonlinearity < 0.0001, the association was approximately linear, and there was a dose-dependent increase in risk with increasing systolic blood pressure from a systolic blood pressure level of 90 mmHg and above (Fig. 3b).

Table 2 Cohort studies of blood pressure and atrial fibrillation
Full size table
Fig. 3
figure 3

Systolic and diastolic blood pressure and atrial fibrillation, linear and nonlinear dose–response analysis

Full size image

Twenty three cohort studies (19 publications, 21 risk estimates) [8, 29, 33, 39, 41, 44, 46, 51, 55, 67, 69, 70, 72,73,74, 77, 79, 81, 82] (333,901 cases, 14,387,470 participants) were included in the analysis of diastolic blood pressure and atrial fibrillation. The summary RR was 1.06 (95% CI: 1.02–1.10, I2 = 92.1%, pheterogeneity < 0.0001) per 10 mmHg increment (Fig. 3c). There was no evidence of publication bias with Egger's test (p = 0.55) or by inspection of the funnel plot (Supplementary Fig. 6). The summary RR ranged from 1.05 (95% CI: 1.00–1.10) when excluding the UK GPRD study [72] to 1.07 (95% CI: 1.03–1.11) when excluding a study at Vanderbilt University [41] (Supplementary Fig. 7). Six cohort studies [39, 55, 67,68,69, 79] were included in the nonlinear dose–response analysis of diastolic blood pressure and atrial fibrillation. There was evidence of nonlinearity (pnonlinearity < 0.0001) with a slightly steeper increase in risk at lower levels of diastolic blood pressure than at higher levels, however, there was an increased risk from a diastolic blood pressure level of around 60 mmHg (Fig. 3d).

Subgroup and sensitivity analyses

There were positive associations between hypertension and risk of atrial fibrillation across all subgroup analyses defined by sex, duration of follow-up, geographic location, number of cases, whether prevalent cases were excluded or not, study quality and adjustment for confounding (and in some cases potentially mediating) factors (including age, education, alcohol, smoking, BMI, physical activity, diabetes, hyperlipidemia, coronary heart disease, heart failure, valvular heart disease, left ventricular hypertrophy, and kidney disease), although the number of studies was small in some subgroups (Table 3). With meta-regression analyses there was some indication of heterogeneity between some subgroups for hypertension, with a stronger association among European studies than studies from the other geographic locations (p = 0.02), and a weaker association among studies with adjustment for smoking (p = 0.03) when compared to those without such adjustment. Further subgroup analyses by ethnicity showed summary RRs of 1.53 (95% CI: 1.29–1.80, I2 = 70.4%, n = 5) for Caucasians [26, 29, 45, 59] and 1.35 (95% CI: 1.16–1.59, I2 = 7.5%, n = 6) for African Americans [26, 29, 45, 46, 59] with no significant heterogeneity between subgroups (p = 0.77) (Supplementary Fig. 8).

Table 3 Subgroup analyses of hypertension, systolic and diastolic blood pressure and atrial fibrillation
Full size table

There was evidence of heterogeneity in the subgroup analysis of systolic blood pressure and atrial fibrillation when stratified by adjustment for education (p = 0.03) and physical activity (p = 0.04) with stronger associations among studies with compared to without such adjustments, however, relatively few studies made such adjustments. For diastolic blood pressure there was heterogeneity in analyses stratified by sex (p = 0.009) and by adjustment for BMI or obesity (p = 0.04), coronary heart disease (p = 0.01), and heart failure (p = 0.009). However, the association was weaker in studies of both sexes combined than in studies among either men or women, and there was no heterogeneity when comparing men with women (and excluding studies in both sexes combined) (Table 3). There was no association in studies that adjusted for BMI or obesity, coronary heart disease, or heart failure, but a positive association in studies that did not make such adjustments (Table 3).

Mean (median) study quality scores were 7.7 (8.0) for the analysis of hypertension, 7.8 (8.0) for systolic blood pressure, and 7.8 (8.0) for diastolic blood pressure.

Discussion

This meta-analysis of cohort studies suggests that persons with hypertension have a 50% increase in the relative risk of developing atrial fibrillation compared to persons without hypertension. There was a 19% increase in the relative risk of atrial fibrillation per 20 mmHg increase in systolic blood pressure and 6% increase in relative risk per 10 mmHg of diastolic blood pressure. Although the test for nonlinearity was significant both for systolic and diastolic blood pressure in relation to atrial fibrillation, the association with systolic blood pressure appeared to be approximately linear, while the association for diastolic blood pressure was nonlinear with a slightly steeper increase in risk at lower levels than at higher levels of diastolic blood pressure. However, there was an increased risk even within what is considered the normal blood pressure range and the lowest risk was observed at a systolic and diastolic blood pressure of 90/60 mmHg, respectively, while there was a 1.8–2.3 fold increase in risk at the high end of systolic and diastolic blood pressure around 180/110 mmHg. Positive associations were observed both in men and women, and among European, American, Asian and Australian studies, however, data from other regions are lacking. In the few studies that reported results stratified by ethnicity, there was a positive association between hypertension and atrial fibrillation among both Caucasians and African Americans. Our findings of an increased risk of atrial fibrillation with higher systolic and diastolic blood pressure are partly consistent with several recent Mendelian Randomization (MR) studies [95,96,97], as well as a randomized open-label trial which found a 54% reduction in risk of new-onset atrial fibrillation among participants allocated to tight vs usual blood systolic blood pressure control (target of < 130 mmHg and < 140 mmHg, respectively) [98], suggesting a possible causal relation between elevated blood pressure and atrial fibrillation. The MR studies reported stronger associations between blood pressure and atrial fibrillation when compared to the current analysis with 17–19% vs. 9% increases in risk of atrial fibrillation per 10 mmHg increase in systolic blood pressure and 25–29% vs. 6% increases in risk of atrial fibrillation per 10 mmHg increase in diastolic blood pressure, respectively. The stronger associations observed in the MR studies could be due to a stronger impact of lifelong elevated blood pressure that may be better captured in the MR studies, and potential overadjustment for intermediate risk factors in some of the observational studies. We did not observe significant differences in the association between hypertension or blood pressure and atrial fibrillation by sex, in contrast to what has been previously observed for cardiovascular disease incidence [99], but consistent with that observed for stroke [100] and cardiovascular disease mortality [99]. This suggests that for the prevention of atrial fibrillation, blood pressure lowering may be equally important among men and women.

Several biological pathways could explain an increased risk of atrial fibrillation in patients with hypertension. Elevated blood pressure increases the risk of coronary heart disease and heart failure [101], conditions that predisposes to atrial fibrillation [29, 34, 47, 48]. Hypertension induces structural remodelling of the left atrium with excessive fibroblast proliferation, and fibroblasts can switch and proliferate to myofibroblasts which have a higher profibrotic potential and also contribute to collagen accumulation [102]. Epidemiological studies have shown that elevated blood pressure predisposes to left ventricular hypertrophy [103,104,105,106], which again increases the risk of atrial fibrillation [17, 20, 29, 34, 41, 51]. It also stimulates apoptosis and inflammation of the cardiomyocytes, leading to fibrosis and left ventricular hypertrophy. Activation of the renin–angiotensin–aldosterone system and autonomic dysregulation are major factors behind these changes. Long-term hypertension can through ventricular thickening, left ventricular hypertrophy and impaired left ventricular systolic-diastolic function increase atrial pressure, ultimately leading to atrial stretch, enlargement and deterioration of atrial contraction [102]. Dysregulation of the autonomic nervous system may also contribute to the development of atrial fibrillation and it has been shown that both sympathetic and parasympathetic overactivation may trigger atrial fibrillation [102].

The present systematic review and meta-analysis has some limitations that need to be discussed. Persons with hypertension often have less healthy lifestyles than persons without hypertension, including higher BMI, less physical activity and they may be more likely to smoke. Several of the included studies adjusted for the most important confounding factors and the results persisted across most subgroup analyses, and we found little evidence of heterogeneity between these subgroups. However, we cannot exclude the possibility that residual confounding could partly explain the results. There was very high heterogeneity in the analyses of hypertension and diastolic blood pressure and moderately high heterogeneity in the analysis of systolic blood pressure and this persisted in many of the subgroup analyses, but there was lower heterogeneity in studies with a longer duration of follow-up. The heterogeneity observed appeared to be more driven by differences in the effect sizes rather than differences in the direction of the association, as all except one study found positive associations between hypertension or systolic blood pressure and atrial fibrillation. Since the studies only had one baseline assessment of hypertension status or blood pressure, regression dilution bias could have attenuated the association between hypertension or blood pressure and risk of atrial fibrillation. Some participants with high blood pressure at baseline may have undergone subsequent treatment for high blood pressure with pharmaceutical medications or lifestyle changes, which would have lowered their blood pressure, but any such effects would most likely have led to conservative estimates of the associations between hypertension and blood pressure and risk of atrial fibrillation. Some of the included studies may also have over-adjusted by including hypertension status and blood pressure in the same models, adjusting rather than stratifying for blood pressure treatment, and/or by adjusting for potentially intermediary conditions such as coronary heart disease, heart failure, valvular heart disease, and left ventricular hypertrophy in the multivariable models. Any further studies might want to adjust for potential confounders and mediators separately to evaluate the impact of both on the observed associations.

Strengths of the present meta-analysis include (1) the cohort design of the included studies (which avoids recall bias and reduces the potential for selection bias), (2) the detailed subgroup and sensitivity analyses, (3) the very large sample size with 14.3–30.5 million participants and 333,000 to 1,080,000 cases providing a more robust estimate of the association between blood pressure and hypertension and risk of atrial fibrillation than most individual studies, and (4) the detailed dose–response analyses. Our findings have important clinical and public health implications as the number of people with hypertension worldwide increased from 594 million in 1975 to 1.13 billion in 2015, mainly due to population growth and ageing, but also due to lifestyle factors [107]. This increase in the number of people with hypertension may at least have partly contributed to increased rates of atrial fibrillation. Routine screening for hypertension and lifestyle interventions to reduce blood pressure that emphasize healthy diets, physical activity, weight control and proper pharmaceutical treatment of hypertension may therefore also reduce the risk of atrial fibrillation as well as other cardiovascular complications.

Conclusion

In conclusion, this meta-analysis suggests that people with hypertension have a 50% increase in the relative risk of developing atrial fibrillation compared to those without hypertension. Increasing systolic and diastolic blood pressure even within the normal range was associated with increased risk and at the high end was associated with a twofold increase in risk. These results strongly support a role of elevated blood pressure in the development of atrial fibrillation.