Abstract
Background
The interrelationships between atrial fibrillation (AF), brain lesions and cognitive function are poorly understood. We aimed to investigate the relationship of AF with brain lesions and cognition.
Methods
We enrolled 1,480 patients with and 959 without AF in a multicenter prospective study (Swiss-AF; NCT02105844). We assessed brain structure, and cognition using the Montreal Cognitive Assessment (MoCA). Brain magnetic resonance imaging (MRI) was performed to assess large non-cortical and cortical infarcts (LNCCI), small non-cortical infarcts (SNCI), white matter hyperintensities (WMH), and microbleeds. Using causal mediation analyses, we investigated the direct (lesion-independent) and indirect (lesion-mediated) effects of AF on cognition.
Results
Mean age in AF patients is 75.0 vs. 74.2 years in no-AF patients, 28.6% vs. 36.9% are female, and comorbidities are comparable. The prevalence of MRI-detected brain infarcts (LNCCI and/or SNCI) is 40.1% in AF patients vs. 24.0% in no-AF patients, adjusted OR (95% CI): 1.78 (1.30; 2.44), p = 0.0003. WMH (Fazekas ≥2) are more prevalent in AF patients (59.2% vs 44.4%), adjusted OR (95% CI): 2.03 (1.50; 2.77), p = 4.6e-06. The mean MoCA score is 25.3 in AF patients and 26.4 in no-AF patients. In mediation analysis, the total effect of AF on cognition is −1.05 MoCA points, decomposed into a direct effect of −0.99 and an indirect, lesion-mediated, effect of −0.06 points.
Conclusions
The prevalence of ischemic brain infarcts and WMH is higher in patients with AF than without AF despite comparable comorbidities. AF is associated with lower cognitive function, primarily through a direct effect rather than mediated by brain lesions.
Subject terms: Cardiology, Neurology
Plain Language Summary
Atrial fibrillation is a common heart rhythm disorder in which there is an irregular heartbeat. It has been associated with damage to the brain and reduced cognitive function, which is the mental processes involved in thinking, learning, and processing information. However, the relationship is poorly understood. We assessed brain structure and cognitive function in people with similar characteristics that either had or did not have atrial fibrillation. We found that people with atrial fibrillation had more brain damage visible by brain imaging and lower cognitive function, compared to people without atrial fibrillation. The brain damage seen did not seem to be the main cause of the lower cognitive function. The main driver of lower cognitive function was most probably the arrhythmia itself.
Krisai et al. compare brain structure and cognitive function in elderly patients with and without atrial fibrillation using brain MRI and cognitive testing. They find that atrial fibrillation is associated with more brain lesions and lower cognitive function, but the cognitive impairment occurs primarily through direct effects of the arrhythmia rather than through brain damage.
Introduction
Atrial fibrillation (AF) and cognitive dysfunction are both highly prevalent, with their incidence continuing to rise in aging populations1–3. Patients with AF have an increased risk of reduced cognitive function, but the mechanisms driving this association remain incompletely understood and cannot be fully explained by clinically overt brain infarcts alone4. It has been shown that patients with AF have, in addition to overt brain infarcts, a high burden of covert brain infarcts on brain magnetic resonance imaging (MRI), even when treated with oral anticoagulation. Moreover, both overt and covert brain infarcts have been similarly associated with reduced cognitivefunction5,6. In addition to AF, patients with reduced cognitive function or dementia often share similar cardiovascular risk factors, such as hypertension, diabetes, and vascular disease7–12. This overlap of risk factors makes it challenging to isolate the independent effect of AF on brain lesions and cognitive function from that of these comorbidities.
The hypothesis underlying this population-based study was that vascular brain lesions are more frequent and the main drivers of reduced cognitive function in patients with AF. To examine this hypothesis, we compared the prevalence of brain lesions and cognitive function in patients with and without AF but similar baseline characteristics and comorbidities. Additionally, we employed causal mediation analysis to decompose the total effect into a natural direct (brain lesion-independent) and a natural indirect (brain lesion-mediated) effect of AF on cognitive function.
We show that AF was associated with an increased burden of brain infarcts, moderate to severe white matter hyperintensities, and lower cognitive function. The effect of AF on cognitive function was mostly not mediated by brain lesions and independent of comorbidities, suggesting that AF may impair cognition through a direct, arrhythmia-related mechanism.
Methods
Patient population
The present study is based on the Swiss Atrial Fibrillation (Swiss-AF) cohort (NCT02105844). The study complies with the Declaration of Helsinki, the study protocol was approved by the local ethics committees (lead ethics committee: Nordwest- und Zentralschweiz (EKNZ); participating ethics committees: Kantonale Ethikkommission Bern, Commission Cantonale d'éthique de la recherche Genève (CCER), Ethikkommission Ostschweiz (EKOS), Comitato etico cantonale Ticino, Commission cantonale d'Éthique de la Recherche sur l'être humain Vaud (CER-VD), Kantonale Ethikkommission Zürich; ID: 2014-067; PB_2016-00793; 2021-00701) and informed written consent was obtained from each patient.
The Swiss-AF study is an ongoing, prospective cohort study at 14 centres in Switzerland investigating the association of AF with brain lesions and cognitive function5,6,13. The main inclusion criteria were previously documented AF and age ≥65 years. A small number of patients aged 45-64 years were enrolled to assess effects of AF on individuals in the active workforce. Exclusion criteria were the inability to give informed consent or secondary AF due to reversible causes. After enrolment of 2,415 patients with AF from 2014 to 2017, we extended the study and enrolled 1,003 patients without AF from 2018 to 2023 (no-AF group). Inclusion criteria for the no-AF group were age ≥65 years, sinus rhythm on the ECG at enrolment and to have no history of AF. About 10% of patients in the no-AF group underwent a Holter-ECG in addition to the mandatory 12-lead ECG at enrolment. Exclusion criteria were the inability to give informed consent, history of AF, atrial flutter or other relevant supraventricular arrhythmias, contraindications for MRI, and diagnosed or suspected dementia. The enrolment of patients with an acute illness was delayed for 4 weeks.
Of the 3,418 enrolled patients, we excluded those <65 years of age (AF group n = 315, no-AF group n = 5), those who had no available MRI data (AF group n = 604, no-AF group n = 24), or cognitive assessment at baseline (AF group n = 11, no-AF group=12) and those with incomplete MRI measurements (AF group n = 5, no-AF group n = 3), leaving 2,439 (71.4%) patients for the current analysis (Supplementary Fig. 1). Main reasons for missing MRI data were cardiac devices and claustrophobia.
Baseline characteristics of excluded patients compared to included patients are shown in Supplementary Table 9.
Clinical variables
At a baseline, in-person study visit, trained study personnel acquired information about patient demographics, prior medical history, interventional and medical treatment and risk factors, using standardized case report forms. After 5 min of rest in a sitting, upright position with legs uncrossed and back supported, blood pressure was measured automated with an appropriately sized cuff and validated device. The mean of three consecutive blood pressure measurements was used in all analyses. AF was categorized into paroxysmal or non-paroxysmal AF according to guideline recommendations at the time of study inception14. Prior stroke was defined as a focal neurological dysfunction with clinical, imaging or pathological evidence of focal infarction due to ischemic, hemorrhagic or undetermined origin. Transient ischemic attack (TIA) was defined as a transient focal neurological dysfunction without evidence of focal infarction.
Brain magnetic resonance imaging
Brain MRI images were acquired by 1.5 or 3 Tesla scanners with a standardized protocol installed on all MR-scanners at local study centres as described previously5.
After local evaluation, all brain MRI data were centrally analyzed by a neuroimaging core lab (Medical Image Analyses Centre, Basel, Switzerland). Blinded expert readers marked and segmented lesions in standardized analyses. Board-certified neuroradiologists confirmed the assessments. Small non-cortical infarcts (SNCI) were defined as lesions ≤20 mm in diameter on FLAIR on axial sections. These lesions could not involve the cortex and had to be consistent with ischemic infarction in the territory of a perforating arteriole located in the white matter, internal or external capsule, deep brain nuclei, thalamus, or brainstem15. There was no differentiation between SNCIs and lacunes. Large non-cortical infarcts were defined with a diameter of >20 mm. Cortical infarcts were defined as lesions of any size on FLAIR involving the cortex. Large non-cortical infarcts and any cortical infarct (LNCCIs) were combined into one category. White matter hyperintensities (WMH) of presumed vascular origin were identified according to STRIVE-2 criteria16. WMH were rated visually using the Fazekas scale, and Fazekas ≥2 was considered to indicate moderate WMH. Cerebral microbleeds (Mb) were defined according to STRIVE-2 criteria and counted on T2*-weighted or susceptibility-weighted imaging. Detailed information on MRI is provided in the supplement.
Cognitive testing
Centrally trained study personnel assessed cognition using the Montreal Cognitive Assessment (MoCA)17. The MoCA evaluates visuospatial abilities, short-term memory, abstraction, language and executive functions divided into 13 individual test sections. Obtainable scores range from 0 to 30 points with higher scores indicating better cognitive function across all domains. One point is added to the total test score if the patient had 12 years or less of formal education.
Statistics and Reproducibility
The full analysis set included all patients with MRI and MoCA scores, as described above. Analyses were based on full outcome measurements of the respective outcome with no outcome imputation performed. Missing covariate measurements were very rare, with up to four missing BMI observations and three patients with unknown smoking status. These were imputed using mean values of continuous variables and the most frequent value for categorical variables. We report continuous baseline characteristics as mean (standard deviation) or median (interquartile range) if strongly skewed, and categorical variables as frequencies (percentages).
In line with our underlying hypothesis, we performed the following analyses: We first investigated the association of AF, AF type and oral anticoagulation with the presence of brain lesions using logistic regression models. At a second step, we used linear regression models to investigate the association of AF with the MoCA score. As MoCA predictors, we included AF and each of the brain lesion types (separate models), or AF and the presence of any of the lesion types. We report effect estimates and 95% confidence intervals, as well p-values for the interaction terms between AF and lesion presence.
Next, we considered that according to our underlying hypothesis brain lesions lie on the causal pathway between AF and cognition, potentially having a mediating effect (Supplementary Fig. 2). Accordingly, as a third step we performed a causal mediation analysis to decompose the potential causal effect of AF to natural direct (brain lesion-independent) and natural indirect (brain lesion-mediated) effects on cognitive function. Again we investigated each brain lesion type separately, as well as the presence of any brain lesion, implementing a weighting-based approach18. Confidence intervals (CI) for the estimated natural effects were calculated via bootstrapping. We used simulations to perform a sensitivity analysis assessing the robustness of our findings to the effect of a potential unmeasured confounder.
Finally, we repeated the analyses of the association between AF and cognitive function in patients without any brain lesions to assess these independently of brain lesions.
All regression analyses were performed unadjusted and adjusted for the following predefined variables: age, sex, BMI, smoking status (active vs. non-smoker), history of hypertension, history of diabetes, education level (basic: ≤6 years, middle: 6 to ≤12 years, advanced: >12 years) and use of oral anticoagulation.
In this exploratory, hypothesis-generating analysis we provided p-values without correction for multiple testing. These should be seen as complementary to the effect sizes, indicating the probability of these effects, or more extreme, under the relevant null-hypothesis, and not as confirmatory with respect to a specific threshold. All statistical analyses were performed using R 4.3.3.
Results
Baseline characteristics for 1,480 patients with AF and 959 patients without AF are shown in Table 1. Mean age was 75.0 vs. 74.2 years in patients with and without AF, 28.6% vs. 36.9% were female, and mean CHA2DS2-VASc score was 3.7 vs. 3.5 points, respectively. Patients with AF more frequently had a history of stroke or TIA (22.1% vs. 13.1%) and a lower rate of coronary heart disease (29.2% vs. 35.0%). Patients with AF had a higher rate of oral anticoagulation (91.6% vs. 8.9 %), but a lower rate of anti-platelet therapy (18.9% vs. 59.1%), than patients without AF. Forty five percent of patients with AF had paroxysmal AF.
Table 1.
Baseline characteristics stratified by presence or absence of atrial fibrillation
| Atrial Fibrillation | No atrial fibrillation | P-value | |
|---|---|---|---|
| n = 1,480 | n = 959 | ||
| Age, years | 75.0 (6.1) | 74.2 (6.0) | 0.002 |
| Female sex | 424 (28.6) | 354 (36.9) | 2.30E-05 |
| BMI, kg/m2 | 27.5 (4.7) | 26.2 (4.3) | 7.82E-12 |
| Highest education | 8.10E-06 | ||
| Basic | 176 (11.9) | 64 (6.7) | |
| Middle | 722 (48.8) | 450 (46.9) | |
| Advanced | 582 (39.3) | 445 (46.4) | |
| Active smoker | 91 (6.1) | 63 (6.6) | 0.74 |
| AF type | n.a. | ||
| Paroxysmal | 666 (45.0) | n.a. | |
| Non-paroxysmal | 814 (55.0) | n.a. | |
| Time since AF diagnosis, years | 3.16 [0.79, 7.96] | n.a. | n.a. |
| Systolic blood pressure, mmHg | 136.2 (18.8) | 136.5 (16.9) | 0.67 |
| Diastolic blood pressure, mmHg | 78.2 (12.1) | 75.9 (9.6) | 9.10E-08 |
| Pulse pressure, mmHg | 58.0 (15.4) | 60.6 (14.5) | 1.31E-05 |
| CHA2DS2-VASc, points | 3.7 (1.6) | 3.5 (1.4) | 0.0031 |
| Diabetes mellitus | 248 (16.8) | 178 (18.6) | 0.28 |
| Sleep apnoea | 186 (12.6) | 98 (10.2) | 0.09 |
| Arterial hypertension | 1064 (71.9) | 643 (67.0) | 0.01 |
| Prior stroke or TIA | 327 (22.1) | 126 (13.1) | 3.59E-08 |
| Prior heart failure | 338 (22.9) | 147 (15.3) | 6.71E-06 |
| Coronary heart disease | 432 (29.2) | 337 (35.2) | 0.002 |
| Peripheral vascular disease | 118 (8.0) | 88 (9.2) | 0.33 |
| Prior renal failure | 297 (20.1) | 142 (14.8) | 0.001 |
| Anticoagulation | 0 | ||
| None | 125 (8.4) | 874 (91.1) | |
| VKA | 585 (39.5) | 13 (1.4) | |
| NOAC | 770 (52.0) | 72 (7.5) | |
| Antiplatelet therapy | 280 (18.9) | 568 (59.2) | 3.51E-92 |
| Beta-blockers | 1052 (71.1) | 297 (31.0) | 5.17E-84 |
| ACE-inhibitors | 449 (30.3) | 257 (26.8) | 0.07 |
| Angiotensin II receptor blockers | 495 (33.4) | 290 (30.2) | 0.1 |
| Aldosterone antagonists | 152 (10.3) | 38 (4.0) | 2.14E-08 |
| Calcium-channel blockers | 350 (23.6) | 238 (24.8) | 0.54 |
Continuous variables are shown as means (standard deviation). Categorical variables are shown as counts (percentages). ACE = angiotensin-converting enzyme; AF = atrial fibrillation; BMI = body mass index; NOAC = non-Vitamin K antagonists; TIA = transient ischemic attack; VKA = Vitamin K antagonists.
AF and brain lesions
The prevalence of brain lesions in both groups is shown in Fig. 1. LNCCI and/or SNCI were found in 594 (40.1%) patients with AF and in 229 (24.0%) patients without AF. In both groups ≥99% had any WMH, but moderate WMH (Fazekas ≥2) were more prevalent in patients with AF (59.2% vs 44.4%). Mb were found in 351 (24.5%) patients with AF and in 256 (27.0%) patients without AF. Frequency of brain lesions stratified by oral anticoagulation status, is provided in Supplementary Table 1.
Fig. 1. Prevalence of brain lesions on MRI in patients with and without atrial fibrillation.
The large bars represent the proportion of any brain lesions in the two groups. In addition, the relevant number of patients (N) and the prevalence of each specific lesion type is shown with individual bars within the groups. AF = atrial fibrillation; LNCCI = large non-cortical and cortical infarcts; Mb = microbleeds; SNCI = small non-cortical infarcts; WMH = white matter hyperintensities.
The odds of infarcts and Fazekas ≥2 were higher in patients with AF, with adjusted odds ratios (aOR) (95% CI) estimated at 1.79 (1.30; 2.45; p = 0.0003) for LNCCI and/or SNCI and 2.03 (1.50; 2.77; p = 4.6e-06) for WMH (Fazekas ≥2). For Mb the aOR was 0.72 (0.51; 1.01; p = 0.06) (Fig. 2). ORs were consistent for paroxysmal and non-paroxysmal AF (Supplementary Table 2). There were no significant interactions of AF and oral anticoagulation for any brain lesion (Supplementary Table 3).
Fig. 2. Summaries of logistic regression models for atrial fibrillation (predictor) and brain lesions (outcomes).
The centre of the error bars indicates the odds ratio. Two-sided p-values of Wald-tests of the odds ratio are given without correction for multiple testing. N = number of patients included in the model. LNCCI = large non-cortical and cortical infarcts; Mb = microbleeds; SNCI = small non-cortical infarcts; WMH = white matter hyperintensities. Adjusting variables included age, sex, BMI, smoking status (active vs. non-smoker), history of hypertension, history of diabetes, education level (basic: ≤6 years, middle: 6 to ≤12 years, advanced: >12 years) and use of oral anticoagulation.
AF and cognitive function
The mean MoCA was 1.1 point lower in patients with AF compared to patients without AF (25.3 vs. 26.4 points) (Fig. 3A). In adjusted linear regression analyses, AF was associated with lower MoCA scores (β [95% CI] −0.93 [−1.33; −0.54; p = 3.44e-06]) (Supplementary Table 4). This difference in cognitive function was consistent across different age groups, and is comparable to the observed difference in MoCA of patients without AF and with a 9.2-year age difference (Supplementary Fig. 3). Importantly, this effect did not vary with age (AF-age interaction p = 0.32). When including only patients with no brain lesions of any type, differences in MoCA between patients with and without AF were consistent, though effect estimates were smaller (Fig. 3B and Supplementary Table 5).
Fig. 3. Cognitive function in patients with AF and patients without AF.
A Summary MoCA scores in patients with AF and patients without AF and B summary conditional on the presence of any lesion identified in MRI. Boxes contain the 25 through 75% quantiles, the thick horizontal lines are the median. Whiskers indicate the most extreme values lying within the box-edge and 1.5 * the interquartile range. Mean values are represented with the crosses and listed below the graph with standard deviation. The number of patients (N) of each group is shown within the plots. Source data is provided in Supplementary Data 2.
In linear regression models, there was a consistent, negative, association of both AF and lesion presence with MoCA score, except for Mb (Table 2). The interactions of AF and lesion were not significant, except for LNCCI (Supplementary Table 6).
Table 2.
Summaries of linear regression models for the association of atrial fibrillation and presence of lesion (predictors; separate models for each lesion type) with MoCA (outcomes)
| Unadjusted | Adjusted | |||||||
|---|---|---|---|---|---|---|---|---|
| Est | 95% CI | p-value | p for interaction | Est | 95% CI | p-value | p for interaction | |
| LNCCI/SNCI | −0.74 | (−0.98; −0.49) | 8.22e-09 | 0.07 | −0.46 | (−0.69; −0.22) | 1.750e-04 | 0.13 |
| AF | −0.96 | (−1.21; −0.72) | 7.54e-15 | −0.89 | (−1.28; −0.50) | 9.89e-06 | ||
| LNCCI | −0.74 | (−1.03; −0.46) | 3.93e-07 | 0.04 | −0.49 | (−0.76; −0.22) | 4.22e-04 | 0.03 |
| AF | −1.02 | (−1.26; −0.78) | 1.09e-16 | −0.92 | (−1.32; −0.53) | 4.12e-06 | ||
| SNCI | −0.70 | (−1.00; −0.40) | 5.88e-06 | 0.23 | −0.42 | (−0.70; −0.13) | 0.004 | 0.44 |
| AF | −1.00 | (−1.24; −0.75) | 1.02e-15 | −0.90 | (−1.29; −0.51) | 7.71e-06 | ||
| Mb | −0.17 | (−0.44; 0.10) | 0.22 | 0.58 | 0.06 | (−0.20; 0.31) | 0.65 | 0.53 |
| AF | −1.08 | (−1.32; −0.84) | 3.26e-18 | −0.91 | (−1.30; −0.51) | 8.51e-06 | ||
| WMH (Faz ≥2) | −0.79 | (−1.02; −0.55) | 5.87e-11 | 0.74 | −0.34 | (−0.57; −0.11) | 0.004 | 0.62 |
| AF | −0.97 | (−1.21; −0.73) | 4.94e-15 | −0.88 | (−1.27; −0.49) | 1.33e-05 | ||
| Any lesion | −0.77 | (−1.03; −0.52) | 2.88e-09 | 0.18 | −0.33 | (−0.57; −0.08) | 0.009 | 0.21 |
| AF | −0.98 | (−1.22; −0.74) | 1.67e-15 | −0.91 | (−1.30; −0.52) | 6.54e-06 | ||
Adjusting variables included age, sex, BMI, smoking status (active vs. non-smoker), history of hypertension, history of diabetes, education level (basic: ≤6 years, middle: 6 to ≤12 years, advanced: >12 years) and use of oral anticoagulation. n = 2434 for LNCCI and/or SNCI; n = 2378 for Mb; n = 2433 for WMH (Fazekas ≥2); n = 2439 for any lesion. AF = atrial fibrillation; LNCCI = large non-cortical and cortical infarcts; Mb = microbleeds; SNCI = small non-cortical infarcts; WMH = white matter hyperintensities.
Mediation analysis of AF and cognition
The results of the causal mediation analysis for the natural direct and natural indirect effect of AF on cognitive function are shown in Fig. 4. The total effects of AF on cognition ranged from −1.02 to −1.05 points of the MoCA score across different lesion types. The natural direct effects of AF were estimated between −0.98 to −1.01 points, while the natural indirect effects, via brain lesions, ranged from 0 to −0.06 points of the MoCA score. For example, when considering the presence of SNCI and/or LNCCI, the total causal effect of AF on MoCA was estimated at −1.02 (−1.39; −0.68), with the natural direct effect of AF estimated at −0.98 (−1.36; −0.63) and the natural indirect effect, via SNCI and/or LNCCI, estimated at −0.04 (−0.09; −0.01). Sensitivity analyses, evaluating the potential impact of unmeasured confounders, showed consistent results. These analyses suggest that even in the presence of a hypothetical strong unmeasured confounder, the natural causal direct effect of AF would still remain substantially larger than the natural indirect effects (Supplementary Table 7).
Fig. 4. Summary of causal mediation analyses for effect decomposition of atrial fibrillation and different types of brain lesions as mediators for MoCA.
The centre of the error bars indicates the estimate. LNCCI = large non-cortical and cortical infarcts; Mb = microbleeds; SNCI = small non-cortical infarcts; WMH = white matter hyperintensities.
Discussion
In this elderly population with a high prevalence of vascular brain lesions on MRI, ischemic brain infarcts were approximately 1.7 times, and moderate to severe white matter hyperintensities about 1.3 times more prevalent in patients with AF compared to patients without AF. These higher rates of brain lesions were found despite comparable comorbidities and the widespread use of oral anticoagulation in AF patients. Patients with AF exhibited lower cognitive function, and AF was independently associated with ischemic brain infarcts, moderate to severe white matter hyperintensities, and lower cognitive function. Mediation analysis suggested that the impact of AF on cognitive function was predominantly independent of brain lesions. These findings suggest that AF may contribute to reduced cognitive function through mechanisms beyond brain lesions.
Ischemic brain infarcts were more frequent in patients with AF compared to patients without AF, despite comparable baseline characteristics and comorbidities. In the ASSERT trial, the lack of a temporal relationship of device-recorded AF episodes and strokes questioned the role of AF as an independent risk factor for stroke, and instead suggested that AF might serve as an indicator of stroke risk through associated comorbidities19. This interpretation could be supported by the high prevalence of brain lesions even in our non-AF cohort and a higher stroke risk in individuals with a high CHA2DS2-VASc score, even in the absence of AF20–22. Alternatively, a driver of brain lesions in our non-AF cohort might have been non-detected subclinical AF, that was previously shown to be present in about 30% of patients without a prior AF diagnosis23,24, though this would be expected to attenuate the currently found association of AF and brain lesions. Recent evidence from a large study involving nearly half a million patients with continuous device monitoring found an increased stroke risk within 5 days of AF episodes lasting >5.5 hours, reinforcing the direct role of AF as a risk factor for stroke25. In our analysis, AF was strongly associated with ischemic brain infarcts and microvascular disease, independent of comorbidities, consistent across different AF types, and unconditional of anticoagulation. Therefore, our results favor the hypothesis that AF itself may be a major risk factor for ischemic brain infarcts and microvascular disease.
In our study, AF was an independent predictor of lower cognitive function, with patients with AF scoring about 1-point lower on MoCA than patients without AF. Although this difference may appear modest, it is comparable to the difference in cognitive function observed in patients with an age difference of about 10 years. This is concerning given the rising prevalence of AF in the aging population. Based on our previous work showing that in patients with AF covert brain lesions lead to a similar decline in cognitive function as overt infarcts5,6, we hypothesized that the effect of AF on cognitive function was driven mainly by MRI-detected brain lesions. While we found that micro- and macrovascular brain lesions do associate with cognition similar to prior studies26, our current findings challenge the hypothesis: Using causal mediation analysis methods, we found that almost all of the association of AF with reduced cognitive function could be explained via the direct path rather than being mediated via any of the investigated brain lesions. The direct path may comprise multiple undifferentiated paths, including unknown mechanisms with potential other, unknown or unmeasured mediators. An effect of AF on cognition, independent of brain lesions, is also supported by our sensitivity analyses, in which we analysed only patients without any brain lesion, and in which results were consistent. The underlying AF related mechanisms remain speculative but may involve blood pressure fluctuations due to irregular AF and varying ventricular cycle lengths, potentially impairing cerebral perfusion27. Among other non-ischemic mechanisms of inflammation and neurodegeneration28, impaired brain glymphatic function has recently been implicated in cognitive decline in patients with AF29. Moreover, AF may be associated with mental health conditions, including depression and/or anxiety, that may accentuate the relation of AF and cognitive function30,31. These potential confounders may be a limitation to the mediation analysis reported here. However, the sensitivity analysis to the presence of unmeasured confounders suggests that this confounding would have to be unreasonably large to change our conclusions.
The higher prevalence of ischemic brain infarcts and lower cognitive function compared to patients without AF, were found in a well anticoagulated AF population. While the cross-sectional design does not allow to draw conclusions on the effect or directionality of anticoagulation on incident brain lesions or cognitive decline over time, we have previously shown that a relevant number of micro- and macrovascular brain lesions and cognitive decline occurred in patients with AF despite a high rate of anticoagulation6. These findings suggest that oral anticoagulation alone does not prevent brain infarcts and loss in cognition in all AF patients. Whether early rhythm control in addition to anticoagulation, as investigated in the EAST-AFNET 4 trial32, attenuates the risk of MRI-detected brain lesions and cognitive decline requires further investigation.
The strengths of our study include the unselected, multicenter cohort with standardized MRI and cognitive assessment, allowing for robust comparisons between patients with and without AF. Limitations of the study population include potential confounding factors, cohort differences and unmeasured variables. The causal mediation analysis itself is prone to limitations in an observational, cross-sectional design, in which the sequential order of AF-diagnosis, development of brain lesions and cognition is not unequivocal – the order we examine is an assumption. This limitation is somewhat contracted in a supplemented analysis performed in the subgroup of patients that exhibited no lesions in MRI, which presented similar results with respect to cognitive differences (Supplementary Table 8). But, residual confounding remains in that analysis, despite our adjustment for known confounders. As randomization to AF or no-AF is not feasible, our approach of comparing patients with and without AF and similar demographics and comorbidities represents a suitable method for addressing these questions. While we screened patients to exclude AF before enrolment, including Holter-ECG in about 10% of the patients in addition to a 12-lead ECG, some patients in the no-AF group might still have had subclinical, non-detected AF. This could have led to an underestimation of the true impact of AF on brain lesions and cognition. We did not have data on the location of brain lesions, brain atrophy and cerebral metabolism, which might have impacted cognitive function. In addition, a possible selection bias lies in the fact that a larger fraction of AF patients did not undergo MRI compared to the non-AF patients; however, the majority of these patients did not undergo MRI due to the presence of a device, and not due to other conditions, suggesting that the bias is likely unsubstantial.
In this population-based study of elderly patients with a high prevalence of vascular brain lesions, AF was associated with an increased burden of brain infarcts, moderate to severe white matter hyperintensities, and lower cognitive function, despite high rates of oral anticoagulation. Importantly, the effect of AF on cognitive function was mostly not mediated by brain lesions and independent of comorbidities, suggesting that AF may impair cognition through a direct, arrhythmia-related mechanism.
Supplementary information
Description of Additional Supplementary files
Acknowledgements
The Swiss-AF study is supported by grants of the Swiss National Science Foundation (grant numbers 33CS30_148474, 33CS30_177520, 32473B_176178, 32003B_197524, and 324730_192394), the Swiss Heart Foundation (FF22036), the Foundation for Cardiovascular Research Basel (FCVR), and the University of Basel.
Author contributions
PK Conceptualization, Methodology, Formal analysis, Data curation, Writing – original draft. SA Methodology, Formal analysis, Writing – original draft. MC Conceptualization, Methodology, Formal analysis, Data curation, Writing – original draft. NR Methodology, Formal analysis, Data acquisition, Writing – original draft. TR Investigation, Data acquisition, Writing – review & editing. ASM Investigation, Data acquisition, Writing – review & editing. JHB Investigation, Data acquisition, Writing – review & editing. PA Investigation, Data acquisition, Writing – review & editing. AA Investigation, Data acquisition, Writing – review & editing. GM Investigation, Data acquisition, Writing – review & editing. RK Investigation, Data acquisition, Writing – review & editing. DS Investigation, Data acquisition, Writing – review & editing. MDV Investigation, Data acquisition, Writing – review & editing. GC Investigation, Data acquisition, Writing – review & editing. G.E. Investigation, Data acquisition, Writing – review & editing. E.H. Investigation, Data acquisition, Writing – review & editing. A.M. Investigation, Data acquisition, Writing – review & editing. R.E.P. Methodology, Data curation, Formal analysis, Writing – original draft. N.R. Investigation, Data acquisition, Writing – review & editing. L.R. Investigation, Data acquisition, Writing – review & editing. M.S. Methodology, Data curation, Formal analysis, Writing – review & editing. C.S. Investigation, Writing – review & editing. P.B. Investigation, Writing – review & editing. C.S.Z. Investigation, Data curation, Writing – review & editing. T.S. Investigation, Data curation, Writing – review & editing. M.D. Investigation, Writing – review & editing. F.M. Investigation, Writing – review & editing. L.H.B. Investigation, Data curation, Writing – review & editing. D.C. Investigation, Funding acquisition, Formal analysis, Writing – review & editing. S.O. Conceptualization, Methodology, Supervision, Project administration, Funding acquisition, Writing – original draft. M.K. Conceptualization, Methodology, Supervision, Project administration, Funding acquisition, Writing – original draft.
Peer review
Peer review information
Communications Medicine thanks Rajiv Mahajan, Rakesh Agarwal and Zuolin Lu for their contribution to the peer review of this work. A peer review file is available.
Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request. To protect participant privacy, this information is not publicly available. Supplementary Data 1 contains the source data for Fig. 3.
Competing interests
Philipp Krisai (PK) reports speaker fees from BMS/Pfizer and Biosense Webster, and research grants from the Swiss National Science Foundation, Swiss Heart Foundation, Foundation for Cardiovascular Research Basel, Machaon Foundation. Michael Kühne (MK) reports grants from Bayer, grants from BMS, grants from Boston Scientific, grants from Daiichi Sankyo, grants from Pfizer, personal fees from Abbott, personal fees from Boston Scientific, personal fees from Daiichi Sankyo, and royalties from Springer (ECG book). Stefan Osswald (SO) reports grants from Swiss National Science Foundation, Swiss Heart Foundation, Foundation for Cardiovascular Research Basel. Felix Mahfoud (FM) has been supported by Deutsche Gesellschaft für Kardiologie (DGK), Deutsche Forschungsgemeinschaft (SFB TRR219, Project-ID 322900939), and Deutsche Herzstiftung. In the past 24 months, Saarland University has received scientific support from Ablative Solutions, Medtronic and ReCor Medical. Until May 2024, FM has received speaker honoraria/consulting fees from Ablative Solutions, Amgen, Astra-Zeneca, Bayer, Boehringer Ingelheim, Inari, Medtronic, Merck, ReCor Medical, Servier, and Terumo. Matthias Schwenkglenks (MS) reports grants from Swiss National Science Foundation, for the conduct of the study; grants from Amgen, grants from MSD, grants from Novartis, grants from Pfizer, grants from Roche, grants and personal fees from BMS and personal fees from Sandoz, all outside the submitted work. David Conen (DC) reports speaker fees from Servier, consulting fees from Trimedics. Giorgio Moschovitis (GM) has received advisory board or speaker’s fees from Astra Zeneca, Bayer, Boehringer Ingelheim, Daiichi Sankyo, Gebro Pharma, Novartis and Vifor, all outside of the submitted work. All other authors declare no competing interest.
Footnotes
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
These authors contributed equally: Stefan Osswald, Michael Kühne.
A list of authors and their affiliations appears at the end of the paper.
Contributor Information
Philipp Krisai, Email: philipp.krisai@usb.ch.
eMBA, for the Swiss-AF Investigators:
Philipp Krisai, Stefanie Aeschbacher, Elisa Hennings, Rebecca E. Paladini, Christian Sticherling, Leo H. Bonati, David Conen, Stefan Osswald, Michael Kühne, Steffen Blum, Désirée Carmine, Ludvig Dahlheim, Corinne Girroy, Andreas Gasser, Ruben Kerler, Vincent Meier, Christine Meyer-Zürn, Pascal B. Meyre, Andreas U. Monsch, Christian Müller, Raffaele Peter, Adrian Schweigler, Gian Völlmin, Priska Zuber, Stefan Osswald, Michael Kühne, Laurent Roten, Drahomir Aujesky, Urs Fischer, Juerg Fuhrer, Simon Jung, Heinrich Mattle, Seraina Netzer, Luise Adam, Carole Elodie Aubert, Martin Feller, Axel Loewe, Elisavet Moutzouri, Claudio Schneider, Tanja Flückiger, Cindy Groen, Lukas Ehrsam, Sven Hellrigl, Alexandra Nuoffer, Damiana Rakovic, Nathalie Schwab, Rylana Wenger, Tu Hanh Zarrabi Saffari, Tobias Reichlin, Nicolas Rodondi, Alain Bernheim, Roger Dillier, Michèle Deubelbeiss, Franz Eberli, Christine Franzini, Daniel Hofer, Raban Jeger, Isabel Juchli, Claudia Liedtke, Samira Murugiah, Jacqueline Nadler, Thayze Obst, Jasmin Roth, Fiona Schlomowitsch, Xiaoye Schneider, Peter Sporns, Katrin Studerus, Noreen Tynan, Dominik Weishaupt, Caroline Wiederkehr, Andreas Müller, Corinne Friedli, Silke Kuest, Karin Scheuch, Denise Hischier, Nicole Bonetti, Alexandra Grau, Jonas Villinger, Eva Laube, Marcel Frick, Stefanie Leuenberger, Franziska Rutz, Jürg-Hans Beer, Angelo Auricchio, Adriana Anesini, Cristina Camporini, Maria Luce Caputo, Rebecca Peronaci, Francois Regoli, Martina Ronchi, Giulio Conte, Roman Brenner, David Altmann, Karin Fink, Michaela Gemperle, Jessica Meier, Peter Ammann, Mathieu Firmann, Sandrine Foucras, Daniel Hayoz, Martine Rime, Michael Kühne, Richard Kobza, Benjamin Berte, Kathrin Bühler, Amir Jadidi, Virgina Justi, Frauke Kellner-Weldon, Melanie Koch, Brigitta Mehmann, Sonja Meier, Myriam Roth, Andrea Ruckli-Kaeppeli, Ian Russi, Kai Schmidt, Mabelle Young, Michael Kühne, Elia Rigamonti, Carlo Cereda, Alessandro Cianfoni, Maria Luisa De Perna, Jane Frangi-Kultalahti, Patrizia Assunta Mayer Melchiorre, Anica Pin, Serena Staglianò, Tatiana Terrot, Luisa Vicari, Giorgio Moschovitis, Georg Ehret, Hervé Gallet, Elise Guillermet, Francois Lazeyras, Karl-Olof Lovblad, Patrick Perret, Philippe Tavel, Dipen Shah, Nathalie Lauriers, Marie Méan, Alessandra Pia Porretta, Sandrine Salzmann, Jürg Schläpfer, Michael Kühne, Andrea Grêt, Jan Novak, Sandra Vitelli, Michael Kühne, Frank-Peter Stephan, Jane Frangi-Kultalahti, Luisa Vicari, Augusto Gallino, Marcello Di Valentino, Matthias Schwenkglenks, Helena Aebersold, Fabienne Foster, Marco Düring, Tim Sinnecker, Anna Altermatt, Michael Amann, Petra Huber, Manuel Hürbin, Esther Ruberte, Alain Thöni, Jens Würfel, Vanessa Zuber, Michael Coslovsky, Pia Neuschwander, Patrick Simon, Ramun Schmid, Christian Baumann, and Nico Ruckstuhl
Supplementary information
The online version contains supplementary material available at 10.1038/s43856-026-01389-w.
References
- 1.Lloyd-Jones, D. M. et al. Lifetime risk for development of atrial fibrillation: the Framingham Heart Study. Circulation110, 1042–1046 (2004). [DOI] [PubMed] [Google Scholar]
- 2.Krijthe, B. P. et al. Projections on the number of individuals with atrial fibrillation in the European Union, from 2000 to 2060. Eur. Heart J.34, 2746–2751 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Prince, M. et al. The global prevalence of dementia: A systematic review and metaanalysis. Alzheimers Dement9, 63–75 (2013). [DOI] [PubMed] [Google Scholar]
- 4.Kalantarian, S., Stern, T. A., Mansour, M. & Ruskin, J. N. Cognitive impairment associated with atrial fibrillation: a meta-analysis. Ann. Intern. Med.158, 338–346 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Conen, D. et al. Relationships of Overt and Silent Brain Lesions With Cognitive Function in Patients With Atrial Fibrillation. J. Am. Coll. Cardiol.73, 989–999 (2019). [DOI] [PubMed] [Google Scholar]
- 6.Kühne, M. et al. Silent brain infarcts impact on cognitive function in atrial fibrillation. Eur. Heart J.43, 2127–2135 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Benjamin, E. J. et al. Independent risk factors for atrial fibrillation in a population-based cohort. The Framingham Heart Study. JAMA271, 840–844 (1994). [PubMed] [Google Scholar]
- 8.Conen, D. et al. Influence of systolic and diastolic blood pressure on the risk of incident atrial fibrillation in women. Circulation119, 2146–2152 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Alonso, A. et al. Blood lipids and the incidence of atrial fibrillation: the Multi-Ethnic Study of Atherosclerosis and the Framingham Heart Study. J. Am. Heart Assoc.3, e001211 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Adelborg, K. et al. Heart failure and risk of dementia: a Danish nationwide population-based cohort study. Eur. J. Heart Fail.19, 253–260 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Lourida, I. et al. Association of Lifestyle and Genetic Risk With Incidence of Dementia. JAMA322, 430–437 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.SPRINT MIND Investigators for the SPRINT Research Group et al. Association of Intensive vs Standard Blood Pressure Control With Cerebral White Matter Lesions. JAMA322, 524–534 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Conen, D. et al. Design of the Swiss Atrial Fibrillation Cohort Study (Swiss-AF): structural brain damage and cognitive decline among patients with atrial fibrillation. Swiss Med. Wkly.147, w14467 (2017). [DOI] [PubMed] [Google Scholar]
- 14.Kirchhof, P. et al. 2016 ESC Guidelines for the management of atrial fibrillation developed in collaboration with EACTS. Eur. Heart J.37, 2893–2962 (2016). [DOI] [PubMed] [Google Scholar]
- 15.Wardlaw, J. M. et al. Neuroimaging standards for research into small vessel disease and its contribution to ageing and neurodegeneration. Lancet Neurol12, 822–838 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Duering, M. et al. Neuroimaging standards for research into small vessel disease-advances since 2013. Lancet Neurol22, 602–618 (2023). [DOI] [PubMed] [Google Scholar]
- 17.Nasreddine, Z. S. et al. The Montreal Cognitive Assessment, MoCA: a brief screening tool for mild cognitive impairment. J. Am. Geriatr. Soc.53, 695–699 (2005). [DOI] [PubMed] [Google Scholar]
- 18.Lange, T., Vansteelandt, S. & Bekaert, M. A simple unified approach for estimating natural direct and indirect effects. Am. J. Epidemiol.176, 190–195 (2012). [DOI] [PubMed] [Google Scholar]
- 19.Brambatti, M. et al. Temporal relationship between subclinical atrial fibrillation and embolic events. Circulation129, 2094–2099 (2014). [DOI] [PubMed] [Google Scholar]
- 20.Mitchell, L. B. et al. Prediction of stroke or TIA in patients without atrial fibrillation using CHADS2 and CHA2DS2-VASc scores. Heart Br. Card. Soc.100, 1524–1530 (2014). [DOI] [PubMed] [Google Scholar]
- 21.Melgaard, L. et al. Assessment of the CHA2DS2-VASc Score in Predicting Ischemic Stroke, Thromboembolism, and Death in Patients With Heart Failure With and Without Atrial Fibrillation. JAMA314, 1030–1038 (2015). [DOI] [PubMed] [Google Scholar]
- 22.Koene, R. J. et al. Relation of the CHA2DS2-VASc Score to Risk of Thrombotic and Embolic Stroke in Community-Dwelling Individuals Without Atrial Fibrillation (From The Atherosclerosis Risk in Communities [ARIC] Study). Am. J. Cardiol.123, 402–408 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Mahajan, R. et al. Subclinical device-detected atrial fibrillation and stroke risk: a systematic review and meta-analysis. Eur Heart J39, 1407–1415 (2018). [DOI] [PubMed] [Google Scholar]
- 24.Svendsen, J. H. et al. Implantable loop recorder detection of atrial fibrillation to prevent stroke (The LOOP Study): a randomised controlled trial. The Lancet398, 1507–1516 (2021). [DOI] [PubMed] [Google Scholar]
- 25.Singer, D. E., Ziegler, P. D., Koehler, J. L., Sarkar, S. & Passman, R. S. Temporal Association Between Episodes of Atrial Fibrillation and Risk of Ischemic Stroke. JAMA Cardiol6, 1364–1369 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Banerjee, G. et al. Cognitive Impairment Before Atrial Fibrillation–Related Ischemic Events: Neuroimaging and Prognostic Associations. J. Am. Heart Assoc.9, e014537 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Gardarsdottir, M. et al. Atrial fibrillation is associated with decreased total cerebral blood flow and brain perfusion. Europace20, 1252–1258 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Krisai, P. et al. Biomarker and cognitive decline in atrial fibrillation: a prospective cohort study. Sci. Rep.15, 12921 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Guo, J. et al. Atrial fibrillation catheter ablation, brain glymphatic function, and cognitive performance. Eur. Heart J. ehaf036 (2025) 10.1093/eurheartj/ehaf036. [DOI] [PubMed]
- 30.von Eisenhart Rothe, A. F. et al. Depression in paroxysmal and persistent atrial fibrillation patients: a cross-sectional comparison of patients enroled in two large clinical trials. Eur. Eur. Pacing Arrhythm. Card. Electrophysiol. J. Work. Groups Card. Pacing Arrhythm. Card. Cell. Electrophysiol. Eur. Soc. Cardiol.16, 812–819 (2014). [DOI] [PubMed] [Google Scholar]
- 31.Al-Kaisey, A. M. et al. Atrial Fibrillation Catheter Ablation vs Medical Therapy and Psychological Distress: A Randomized Clinical Trial. JAMA330, 925–933 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Kirchhof, P. et al. Early Rhythm-Control Therapy in Patients with Atrial Fibrillation. N. Engl. J. Med. (2020) 10.1056/NEJMoa2019422. [DOI] [PubMed]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Description of Additional Supplementary files
Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request. To protect participant privacy, this information is not publicly available. Supplementary Data 1 contains the source data for Fig. 3.