Coffee Consumption Correlates With Better Cognitive Performance in Patients With a High Incidence for Stroke

Abstract

Background

Atrial fibrillation is an independent risk factor for the development of cognitive impairments. Regular coffee consumption has shown cognitive benefits in healthy individuals. Whether regular consumption reduces cognitive decline in vulnerable patients is controversial. We investigated the association in elderly people with atrial fibrillation.

Methods and Results

Daily coffee consumption was assessed using a structured nutrition questionnaire, and cognitive function was evaluated by a detailed neurocognitive‐test‐battery, including the Montreal Cognitive Assessment, Trail‐Making Test, semantic fluency, and Digit‐Symbol‐Substitution Test. The cognitive construct score combines all neurocognitive tests mentioned and provides an overall cognitive performance indicator. Hs‐CRP (high‐sensitivity C‐reactive protein) and IL‐6 (interleukin‐6) were measured to explore an association with inflammation. Results were estimated using linear mixed‐effects‐models with detailed adjustments for confounders. The <1 cup/day consumers (reference group) reached a cognitive construct score of −0.24 (95% CI, –0.27 to –0.16), and the group with the highest consumption (>5 cups/day) was at −0.10 (95% CI, –0.10 to 0.04; p=0.048). Montreal Cognitive Assessment score in the reference group was 24.58 (95% CI, 24.58–25.32); the group with the highest intake achieved 25.25 (95% CI, 24.98–26.85; p=0.163). Inflammatory markers decreased with higher coffee consumption (hs‐CRP with 5 compared with <1 cup/day by factor 0.78 [95% CI, 0.54–1.13], p= 0.188, IL‐6 significantly by factor 0.73 [95% CI, 0.57–0.95], p=0.017).

Conclusions

Coffee consumption in patients with atrial fibrillation may be associated with improved cognitive performance and reduced inflammatory markers. Further research is needed to confirm these findings and to consider implementation in dietary counseling for atrial fibrillation management.

Registration

URL: https://www.clinicaltrials.gov; Identifier: NCT02105844.

Nonstandard Abbreviations and Acronyms

CoCo

cognitive construct

DSST

Digit Symbol Substitution Test

MoCA

Montreal Cognitive Assessment

SF

semantic fluency

Swiss‐AF

Swiss Atrial Fibrillation Cohort Study

TMT

Trail Making Test

Clinical Perspective.

What Is New?

• Coffee consumption associates with better cognitive performance in patients with atrial fibrillation.

What Are The Clinical Implications?

• Based on our findings and the current literature, clinicians should not discourage regular coffee consumption in patients with atrial fibrillation.

Atrial fibrillation (AF) is the most frequent type of cardiac arrhythmia, affecting 5% of the population >65 years and increasing exponentially thereafter. ¹ AF results in a 5‐fold increased risk of thromboembolic stroke. Furthermore, it represents an independent risk factor for developing mild cognitive impairment and dementia with odds ratios of 2.3 to 5.8 respectively. ² , ³ , ⁴ , ⁵ Deficits affecting visual and verbal memory, executive functions, and cognitive inflexibility have been reported. ⁶ , ⁷ Secondary prophylaxis with anticoagulants reduced incidence of strokes and dementia. ³ , ⁸ , ⁹ AF is not only associated with higher incidence of vascular cognitive impairment ¹⁰ and increased size and number of strokes, but also with an increased risk for Alzheimer disease. ⁸ , ¹¹ As reported by neuropathological and epidemiological studies, possible underlying mechanisms include an altered cerebral blood flow, an imbalance of pro‐ and antithrombotic factors, and different levels of trophic factors, such as the brain‐derived neurotrophic factor and vascular endothelial‐derived growth factor. ¹²

Epidemiologic data suggest protective effects of regular coffee consumption against cognitive decline, especially in elderly people. ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ Others report no significant effects. ¹⁸ An association with a lower stroke incidence has been described. ¹⁹ , ²⁰ Beneficial effects were attributed to caffeine and other active ingredients including trigonelline, melanoidins, magnesium, and vitamin B₃ (niacin). ¹³ Underlying mechanisms that have been postulated include the reduction of oxidative stress, ²¹ reduction of proinflammatory mediators (TNFα [tumor necrosis factor alpha], IL‐6 [interleukin‐6], IL‐10, IL‐12), ²² , ²³ increased neuroplasticity ²⁴ attenuation of blood–brain‐barrier leakage, ²⁵ and the reduction of Aβ‐amyloid. ²⁴

We therefore hypothesized, that coffee consumption may contribute to attenuating cognitive decline in patients with AF. We chose the model of AF because it is a frequent and relevant condition and is associated with a high prevalence of mild cognitive impairment and dementia; this renders this population particularly well suited for the analysis of such a critical endpoint. Furthermore, we also conducted an exploratory analysis of the inflammatory parameters hs‐CRP (high‐sensitivity C‐reactive protein) and IL‐6 and their correlation with different levels of coffee consumption because, as cited previously, earlier reports have considered underlying anti‐inflammatory mechanisms. ²² , ²³

METHODS

Data Availability

Because of the sensitive nature of the data collected for this study, requests to access the data set from qualified researchers trained in human subject confidentiality protocols may be sent to the study coordinator of the Swiss‐AF (Swiss Atrial Fibrillation Cohort Study) at rebecca.paladini@usb.ch.

Ethics Approval

We conducted the analyses as part of the Swiss‐AF study cohort (ClinicalTrials.gov Identifier: NCT02105844), which was approved by the ethics commission of northwestern and central Switzerland. All patients provided informed consent. Detailed information on the design of the cohort has been published earlier. ²⁶

Study Population

The Swiss‐AF study is an ongoing, observational, prospective, 14‐center‐cohort study, which enrolled 2415 patients with AF between 2014 and 2017 in Switzerland. The study aims to provide novel insights into structural and functional brain alterations in patients with AF. Eligible patients had to be at least 65 years of age (except for 250 patients <65 years of age) and to show a history of documented AF. Patients with only short episodes of reversible (<30 min) forms of AF (eg, due to infection or surgery) and acute illness within the past 4 weeks were not included, as well those with inability to provide informed consent.

Assessment Baseline Characteristics and Coffee Consumption Habits

Baseline characteristics were assessed using structured questionnaire (Data S1), which has been previously validated. ²⁶ Caffeinated coffee consumption of the past 12 months was assessed and indicated as “never or less than once per month.” “1 to 3 cups per month,” “1 per week,” “2 to 4 per week,” “5 to 6 per week,” “once daily,” “2 to 3 daily,” “4 to 5 daily,” or “6+ daily.” To ensure comparability with previous studies, we simplified the exposure groups to <1 cup per day, 1 cup per day, 2 to 3 cups per day, 4 to 5 cups, and >5 cups per day.

Neurocognitive Assessments

Instructed study personnel performed standardized neurocognitive assessments. First, the Montreal Cognitive Assessment (MoCA) questionnaire ²⁷ assessed visuospatial and executive functions, confrontation naming, memory, attention, language, and abstraction abilities. The MoCA test is a widely used tool to screen for cognitive impairment. ²⁷ Patients can reach a maximal score of 30 points.

Information about psychomotor speed and executive functions were obtained using the Trail‐Making Test (TMT) A and B, ²⁸ showing the amount of performed connections per second. Semantic fluency (SF), based on semantic memory, executive functions, and the assessment of phonetic fluency, was assessed by the test of animal fluency for 1 minute. ²⁹ Finally, the Digit Symbol Substitution Test (DSST) assessed the processing speed, visuomotor coordination, and attention. Scores range from 0 to 90 ³⁰ and indicate the amount of substituted symbols per minute.

Finally, the cognitive construct (CoCo) score, a global variable reflecting one's cognitive abilities as measured by the test battery described, was calculated. This score allows indicating each patient's cognitive performance as a single score. ³¹ It was derived as a factor score based on a principal component analysis conducted within the Swiss‐AF cohort study. It showed larger effect size as compared with the MoCA score in relation to relevant clinical variables, ³¹ thus, allowing an increased measurement sensitivity.

In addition, we performed age‐ (<80 versus >80 years of age) and sex‐stratified subanalyses for the CoCo and MoCA scores.

Because a potential concomitant depression might introduce bias (eg, pseudodementia ³² ), we assessed depressive symptoms using the geriatric depression scale. A score of 10 or more points indicates a possible concomitant depression ³³ and was handled as a confounding factor.

Blood Samples

Blood was collected by standard operating procedures. Plasma and whole blood aliquots were immediately processed, divided into cryotubes, and stored at −80 °C in the centralized biobank at the University Hospital Basel. ²⁶ Inflammatory markers (the hs‐CRP and the IL‐6) were measured by standard ELISA procedures.

End Points

Primary outcomes were the CoCo and MoCA scores; secondary endpoints included TMT‐A and B, DSST, and SF. Exploratorily, we determined the inflammatory biomarkers hs‐CRP and IL‐6.

Statistical Analysis

Continuous variables are reported as mean±SD or median and interquartile ranges for hs‐CRP and IL‐6. Categorical variables are given as frequencies and percentages. The chi‐square test was used for categorical variables. For continuous variables, the ttest was applied if they were normally distributed, and the Kruskal–Wallis test was used for nonnormally distributed variables. Fulfilling all assumptions for linear regression (see Figure S1A through S1H), the association between coffee consumption and cognitive performance was estimated using a linear mixed effects model, where study centers were used as random effects and following variables as covariates: current coffee consumption, age, sex, education, body mass index (BMI), smoking status, physical activity (as a categorical variable), presence of depression, arterial hypertension, medication with anticoagulants, diabetes type 2 and history of stroke and for CoCo and MoCA hs‐CRP and IL‐6 (for better visualization of the different interaction, we added a directed acyclic graphs, see Figure S2). Additional calculations using robust variance estimators are included within Table S1. P values were adjusted for multiple comparisons. We tested for possible interaction of age and sex with coffee consumption (see Figure S3A and S3B). Within the linear mixed effects model, the beta coefficient (β) represents the change in the cognitive test score for an increase in coffee consumption group (eg, from the <1 cup/day group to the 1 cup/day group). Polynomial contrasts were employed to estimate linear trend of coffee consumption and cognitive decline. Only patients with complete records were included (Figure S4). 95% profile CIs were computed for all estimates.

The inflammatory parameter distributions were skewed. To reduce the influence of extreme values, concentrations were log‐transformed (Figure S5). We added 0.001 to the IL‐6 values to allow the logarithmic transformation.

To isolate possible confounders, we conducted several sensitivity analyses, including an analysis solely with patients without depression, patients with and without hypertension, and a model in which we adjusted for systolic blood pressure (see Tables S2–S5).

All statistical analyses were performed in R, version 4.0.2.

RESULTS

Baseline Data

From the 2415 patients of the Swiss‐AF cohort, baseline data on 2413 patients (99.8%) are fully available (Table 1). Two patients with missing coffee consumption were excluded from the analysis (see Figure S4).

Table 1.

Baseline Characteristics and Observed Outcomes by Coffee Consumer Groups

Variables	Overall	<1 cup	1 cup	2–3 cups	4–5 cups	>5 cups	P value
Baseline characteristics
Number of patients, n (%)	2413 (100)	491 (20.3)	475 (19.7)	1165 (48.3)	240 (9.9)	42 (1.7)
Patient's sex male, n (%)	1751 (72.6)	338 (68.8)	316 (66.5)	858 (73.6)	201 (83.8)	38 (90.5)	<0.001
Age at baseline, y, mean±SD	73.25±8.42	73.77±9.22	73.96±8.13	73.23±8.10	71.96±8.42	67.15±7.64	<0.001
Level of education, n (%)							0.054
Basic	288 (11.9)	64 (13.1)	76 (16.0)	120 (10.3)	22 (9.2)	6 (14.3)
Middle	1197 (49.6)	227 (46.3)	224 (47.2)	601 (51.6)	124 (51.7)	21 (50.0)
Advanced	926 (38.4)	199 (40.6)	175 (36.8)	443 (38.1)	94 (39.2)	15 (35.7)
Body mass index, kg/m2, mean±SD	27.68±4.79	27.48±4.76	27.74±4.79	27.70±4.84	27.56±4.37	29.41±5.84	0.159
Physically active, n (%)	1112 (46.1)	226 (46.0)	208 (43.8)	550 (47.2)	116 (48.3)	12 (28.6)	0.122
Active smoker, n (%)	175 (7.3)	29 (5.9)	24 (5.1)	87 (7.5)	25 (10.4)	10 (23.8)	<0.001
History of stroke, n (%)	318 (13.2)	58 (11.8)	59 (12.4)	153 (13.1)	42 (17.5)	6 (14.3)	0.287
Arterial hypertension, n (%)	1683 (69.7)	347 (70.7)	339 (71.4)	812 (69.7)	159 (66.2)	26 (61.9)	0.492
Diabetes type 2, n (%)	411 (17.0)	72 (14.7)	73 (15.4)	217 (18.6)	40 (16.7)	9 (21.4)	0.233
Geriatric depression scale category, n (%)							0.002
Normal	2290 (95.1)	468 (95.5)	444 (94.1)	1117 (95.9)	222 (92.5)	39 (92.9)
Borderline	102 (4.2)	20 (4.1)	18 (3.8)	46 (3.9)	16 (6.7)	2 (4.8)
Depression	17 (0.7)	2 (0.4)	10 (2.1)	2 (0.2)	2 (0.8)	1 (2.4)
Use of anticoagulants n (%)							0.229
No	233 (9.7)	47 (9.6)	35 (7.4)	116 (10.0)	29 (12.1)	6 (14.3)
Yes	2180 (90.3)	444 (90.4)	440 (92.6)	1049 (90.0)	211 (87.9)	36 (85.7)
Primary outcomes
CoCo to median [IQR]	‐0.02 [−0.36 to 0.33]	−0.10 [−0.43 to 0.30]	−0.05 [−0.38 to 0.25]	−0.01 [−0.32 to 0.35]	0.07 [−0.36 to 0.46]	0.20 [−0.12 to 0.51]	<0.001
MoCA to mean±SD	25.36±3.13	25.06±3.29	25.30±3.05	25.46±3.11	25.48±3.11	26.15±2.65	0.069
Secondary outcomes
TMT A (connections per second) median [IQR]	0.50 [0.38 to 0.66]	0.48 [0.37 to 0.63]	0.49 [0.37 to 0.63]	0.50 [0.38 to 0.66]	0.53 [0.40 to 0.71]	0.59 [0.45 to 0.75]	0.004
TMT B (connections per second) to median [IQR]	0.20 [0.14 to 0.27]	0.18 [0.12 to 0.26]	0.19 [0.19 to 0.26]	0.20 [0.14 to 0.27]	0.22 [0.14 to 0.30]	0.22 [0.18 to 0.33]	<0.001
DSST (substitutions per minute) to mean±SD	43.62±14.27	42.63±15.03	42.28±13.63	43.94±13.76	45.71±15.67	49.77±14.89	0.001
SF to mean±SD	18.87±5.41	18.45±5.59	18.10±5.03	19.21±5.41	19.55±5.65	19.24±4.87	<0.001
Inflammatory parameters
High‐sensitivity C‐reactive protein (mg/L) to median (IQR)	1.71 (0.86 to 4.00)	1.85 (0.96 to 4.19)	1.93 (0.90. 4.22)	1.67 (0.81 to 3.93)	1.50 (0.76 to 3.75)	1.62 (0.75 to 3.80)	0.132
Interleukin‐6 pg/mL to median (IQR)	3.62 (2.33 to 5.93)	3.66 (2.30 to 6.10)	4.00 (2.46 to 6.52)	3.52 (2.33 to 5.57)	3.39 (2.13 to 5.94)	3.58 (2.74 to 4.83)	0.054

	Effect estimate	95% CI	P value
CoCo score [respect to reference group (<1 cup to CoCo=−0.24)]
1 cup	0.02	−0.03 to 0.08	0.427
2–3 cups	0.06	0.01 to 0.10	0.026
4–5 cups	0.11	0.04 to 0.18	0.002
>5 cups	0.14	0.00 to 0.27	0.074
MoCA score [respect to reference group (<1 cup=MoCA=24.58)]
1 cup	0.37	0.00 to 0.74	0.090
2–3 cups	0.38	0.07 to 0.69	0.029
4–5 cups	0.39	−0.06 to 0.84	0.140
>5 cups	0.68	−0.26 to 1.61	0.196
TMT‐A score [respect to reference group (<1 cup=0.47 connections per second)]
1 cup	0.00	−0.02 to 0.03	0.793
2–3 cups	0.02	0.00 to 0.04	0.253
4–5 cups	0.03	0.00 to 0.06	0.156
>5 cups	0.02	−0.04 to 0.08	0.667
TMT‐B score [respect to reference group (<1 cup=0.17 connections per second)]
1 cup	0.00	−0.01 to 0.01	0.766
2–3 cups	0.01	0.00 to 0.02	0.178
4–5 cups	0.02	0.01 to 0.04	0.007
>5 cups	0.02	−0.01 to 0.05	0.243
SF Score [respect to reference group (<1 cup=17.18 words per minute)]
1 cup	−0.01	−0.66 to 0.64	0.978
2–3 cups	0.74	0.19 to 1.28	0.020
4–5 cups	0.94	0.14 to 1.73	0.041
>5 cups	0.47	−1.14 to 2.09	0.628
DSST Score [respect to reference group (<1 cup=37.88 symbols in 120 second)]
1 cup	0.75	−0.81 to 2.30	0.386
2–3 cups	1.18	−0.11 to 2.48	0.091
4–5 cups	2.77	0.87 to 4.66	0.008
>5 cups	4.89	0.95 to 8.82	0.023

Abbreviations: CoCo indicates cognitive construct; DSST, Digit Symbol Substitution Test; MoCA, Montreal Cognitive Assessment; SF, semantic fluency; and TMT, trail‐making test.

Baseline characteristics of the patients are summarized in Table 1. Most prevalent quantity of coffee consumption was 2 to 3 cups/day (n=1165, 48.3%, Figure S6). Mean age was 73 years of age, 72.6% of the patients were male, and 49.6% had an average level of education, corresponding to a high school diploma or a completed apprenticeship. All coffee consumer groups showed a mean BMI of >25 kg/m² (Table 1). Prevalence of previous strokes ranged from 11.8% to 17.5% with highest rate in the group with 4 to 5 cups/day (Table 1). Arterial hypertension was the most common cardiovascular risk factor (overall 69.7%) with lowest prevalence in the highest coffee consumer group (61.9%). Diabetes was present in 17.0% of patients, with highest rate in the >5 cups/day group (21.4%). Highest percentage of active smokers was recorded in the >5 cups/day group (23.8%, Table 1). Most participants (95.1%) reached normal (<5 points) geriatric depression scale scores.

Primary Outcomes

CoCo‐Score

The distribution of the CoCo scores is shown in Figure 1A and 1B and Table 1. All coffee consumer groups showed higher CoCo scores compared with the comparator (<1 cup/day), with statistical significance for the 2to 3 and 4 to 5 cups/day groups. The highest estimated CoCo score was reached by the group >5 cups/day. Using polynomial contrasts, the linear trend estimate on CoCo score was significant with 0.11 (95% CI, 0.03–0.20; P=0.01) additional point per higher coffee consumption group. We could not observe any interactions age, sex, and coffee consumption (see Figure S3A).

Figure 1. Primary outcomes.

CoCo score as aggregated global performance parameter. A, Boxplot visualizing unadjusted scores subdivided in different coffee consumer groups. B, Estimates of adjusted CoCo scores indicating differences between coffee consumer groups and the reference group (<1 cup of coffee per day) indicating higher cognitive performance with higher coffee consumption and showing the P value for trends. C, Boxplot visualizing unadjusted MoCA in different coffee consumer groups. D, Estimates of adjusted MoCA scores and P value for trends. The data provide the evidence of a persistent significance of the caffeine effect on cognitive function after extensive adjustments for the following confounders: Current coffee consumption, age, sex, education, body mass index, smoking status, physical activity, presence of depression, arterial hypertension, medication with anticoagulants, diabetes type 2, and history of stroke, as well as high‐sensitivity C‐reactive protein, and interleukin‐6 levels. CoCo indicates cognitive construct; and MoCA, Montreal cognitive assessment score.

Sex and age‐stratified subanalyses indicated significant positive associations for men with 2 to 3, 4 to 5, and >5 cups/day. Men <80 years of age were significantly associated with 2 to 3, 4 to 5, and >5 cups/day, whereas men >80 years associated positively with 1 cup/day (Figure S7 and Table S6).

MoCA‐Scores

The MoCA scores as per coffee consumption group are shown in Figure 1C and 1D and Table 1. All coffee consumer groups showed increased estimated MoCA scores compared with the comparator with statistical significance for the 2 to 3 cups/day group. Highest estimated MoCA score was reached by the group with >5 cups/day. Using polynomial contrasts, the linear trend estimate was 0.43 (95% CI, –0.17 to 1.05; P=0.16) additional point per higher coffee consumption group. With exception of the male >5 cups/day group, we could not observe any interaction between age, sex, and coffee consumption (see Figure S3B).

Age‐ and sex‐stratified subanalyses indicated significant benefit with 1 and 2 to 3 cups/day for men >80 years of age and with 2 to 3 cups/day for men <80 years of age (see Table S7 and Figure S8). In women, no group reached significant results.

Secondary Outcomes

Digit Symbol Substitution Test

All consumer groups showed higher estimates compared with the reference group (Figure 2). We observed the highest association with 4 to 5 and >5 cups/day (Table 1), which reached significance. Using polynomial contrasts, the linear trend estimate of 3.77 was observed (95% CI, 1.17–6.29; P≤0.0041).

Figure 2. Same pattern in the DSST and the SF scores; particularly the DSST demonstrates a strong and dose‐dependent association.

Secondary outcomes: Cognitive subtests performed within the Swiss‐AF study. A, Boxplot visualizing unadjusted DSST scores. B, Estimates of adjusted DSST scores and P value for trends. C, Boxplot of unadjusted SF scores in different coffee consumer groups. D, Estimates of adjusted SF scores and P value for trends. The data collectively indicate a significant effect of coffee consumption in the DSST and SF test, which remained significant after adjustments in the DSST and in the larger subgroups in the SF test. Adjusting covariates: Current coffee consumption, age, sex, education, body mass index, smoking status, physical activity, presence of depression, arterial hypertension, medication with anticoagulants, diabetes type 2, and history of stroke, as well as high‐sensitivity C‐reactive protein and interleukin‐6 levels. DSST indicates Digit Symbol Substitution Test; SF, semantic fluency; and Swiss‐AF, Swiss Atrial Fibrillation Cohort Study.

Semantic Fluency

The distribution of the SF scores is shown in Figure 2. We observed highest performance with 4 to 5 cups/day. After adjustment, statistically significant difference compared with the reference group was found for the 2 to 3 and 4 to 5 cups/day group (Table 1). Using polynomial contrasts, linear trend estimate on SF was 0.60 (95% CI, –0.45 to 1.65; P=0.27).

Trail‐Making Test

With increasing coffee consumption, higher estimates in all coffee consumption groups compared with the comparator were observed (Figure 3 and Table 1). Significant difference was observed for the TMT‐B test with 4–5 cups/day. Using polynomial contrasts, we observed a linear trend estimate in TMT‐A of 0.02 (95% CI, −0.02 to 0.06; P=0.29) and in TMT‐B of 0.02 (95% CI, 0.00–0.04; P=0.05).

Figure 3. A similar pattern can be observed in the TMT‐A and TMT‐B test, which reach significance with 4 to 5 cups in TMT‐B.

Secondary outcomes: Boxplot of unadjusted (A) TMT‐A and (B) TMT‐B scores in different coffee consumption groups and P value for trends. Estimates indicating differences of (C) TMT‐A and (D) TMT‐B between coffee consumer groups and reference groups (<1 cup of coffee per day) and P value for trends, adjusted for the aforementioned variables. The data collectively shows a significant association in TMT‐A and ‐B and its persistence in TMT‐B in the highest coffee consumption groups after extensive adjustments. Adjusting covariates: Current coffee consumption, age, sex, education, body mass index, smoking status, physical activity, presence of depression, arterial hypertension, medication with anticoagulants, diabetes type 2, and history of stroke, as well as high‐sensitivity C‐reactive protein and interleukin‐6 levels. TMT indicates Trial Making Test.

Inflammatory Marker

The inverse association of the log‐transformed hs‐CRP and IL‐6 levels with coffee consumption is shown in Figure 4A through 4D, Table 1, and Table S8. Lower hs‐CRP and IL‐6 concentrations were found with increasing coffee consumption.

Figure 4. Inflammatory parameters.

Inflammatory parameters in association to coffee consumption. A, Boxplot visualizing unadjusted log(hs‐CRP) concentration in different coffee consumer groups. B, Estimates of adjusted log(hs‐CRP) concentrations and P value for trends. C, Boxplot of unadjusted log(IL‐6) concentrations. D, Estimates of adjusted log(IL‐6) concentrations and P value for trends. Association between dose‐dependent coffee consumption and preserved cognitive function in patients with AF: As possible protective mechanism, an anti‐inflammatory effect is detectable. Further, in literature described effects are listed. Using polynomial contrast, the linear trend estimate of coffee consumption on log(hs‐CRP) was −0.18 (95% CI, –0.43 to 0.06; P=0.14) and on log(IL‐6) –0.2 (95% CI, –0.37 to −0.04; P=0.02). Adjusting covariates: Current coffee consumption, age, sex, education, body mass index, smoking status, physical activity, presence of depression, arterial hypertension, medication with anticoagulants, diabetes type 2, and history of stroke. AF indicates atrial fibrillation; hs‐CRP, high‐sensitivity C‐reactive protein; and IL‐6, interleukin‐6.

Backtransforming the estimates to the original scale, compared with <1 cup/day, drinking 1 cup/day multiplies the hs‐CRP value by 0.96 (95% CI, 0.83–1.11; P=0.629), with >5 cups/day by 0.78 (95% CI, 0.54–1.13; P=0.458) (Table S8). Using polynomial contrast, the linear trend estimate was −0.18 (95% CI, –0.43 to 0.06; P=0.14).

For the IL‐6 values, coffee consumer group with 2, 4 to 5, and >5 cups/day associated with a lower concentration (see Figure 4B and 4D and Table S8). Backtransforming the estimates to the original scale, drinking 1 cup/day multiplies the IL‐6 value by 1.01 (95% CI, 0.91–1.12; P=0.819), and with >5 cups/day by 0.73 (95% CI, 0.57–0.95; P=0.050) compared with the reference group. Using polynomial contrasts, the linear trend estimate on log(IL‐6) was −0.20 (95% CI, –0.37 to –0.04; P=0.02).

Sensitivity Analysis

In the sensitivity analysis with patients exclusively without depression, we observed for some coffee consumer groups higher estimates for MoCA, DSST, and CoCo scores, which partly reached statistically significant levels (eg, the >5 cups/day group for CoCo and DSST, see Table S5A and S5B).

Stratifying by hypertension status, we observed higher performances across all tests in the group without hypertension compared with the group with hypertension, however not reaching the same statistical significance levels in the group without hypertension (see Tables S3 and S4). Regarding the inflammatory parameters, we observed an anti‐inflammatory effect across all consumer groups in the cohort with hypertension, whereas in patients without hypertension, this effect was limited to the consumers of 4 to 5 and >5 cups/day (see Tables S3 and S4). Finally, after adjusting for systolic blood pressure, we did not observe any relevant changes in value or significance level for any parameter.

DISCUSSION

In this study of patients with AF, our main finding is the association between daily coffee consumption and improved cognitive performance. The association is robust, dose‐dependent, and detectable starting with 1 cup of coffee.

The findings are documented using a panel of independent and well‐validated cognitive tests. Most findings on cognitive function are linear. The association persists even after detailed adjustment for many confounding factors. The biological importance is underlined by the finding that a 1‐point difference in the MoCA score is equivalent to an average 10‐year age difference or the presence of diabetes or arterial hypertension respectively. ³⁴ Accordingly, the effect size of >5 cups/day is comparable with a calculated age difference of 6.7 years.

In line with previous studies, our results highlight that the growing AF population is a particularly vulnerable population for the development of cognitive impairments ³ , ⁴ , ⁵ , ²⁷ and prevalence is high because no group showed a normal, mean MoCA score (>26 points ²⁷ ).

Our findings showed that the highest coffee consumers reached the highest estimated MoCA score (25.25), despite exhibiting the highest risk profile (highest BMI, active smokers, which is known to be associated with coffee consumption ³⁵ ), incidence of diabetes, lower educational level and physical activity. Accordingly, this group showed second highest prevalence for stroke. Contrary, prevalence of arterial hypertension remarkably decreased. In line with our findings, a recent systematic review reported an inverse association between coffee consumption and arterial hypertension. ³⁶ However, with 42 patients (corresponding to 1.74% of the study population), this group of consumer represents a minority, and results should be interpretated with caution. Nonetheless, considering the linear trend estimates and the results of the other coffee consumer groups, the observed scores appear plausible.

Considering the subgroup analysis without patients suffering from depression, we observed some higher estimates, partly reaching statistical significance (eg, for CoCo and DSST), thus underlying the negative effects of such a condition. Stratifying the patients with versus without hypertension, we observed indeed higher performances among patients without hypertension. Nevertheless, an improvement in performances with higher coffee consumption could be observed for both groups. Considering the inflammatory parameters, we observe higher hs‐CRP and IL‐6 concentrations in patients with hypertension and, interestingly also, higher anti‐inflammatory effects with coffee consumption. Thus, hypertension appears to be a major driver for the development of cognitive impairment. Adjusting for systolic blood pressure, we did not observe any relevant changes of the estimates.

The literature on effects of a regular coffee consumption on cognitive decline is controversial: Systematic reviews, ¹⁴ , ³⁷ cross‐sectional, ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ , ⁴² , ⁴³ and longitudinal studies ¹⁵ , ²³ , ⁴⁴ , ⁴⁵ , ⁴⁶ , ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ , ⁵² reported positive effects, even though with highly variable effect sizes. In contrary, the Finnish Twin Cohort Study and the HAAS (Honolulu‐Asia Aging Study), both with a population aged between 46 and 52 years, ¹⁸ , ⁵⁰ did not record any significant effects. However, autopsies from 418 brains from the HAAS study showed a lower prevalence of Alzheimer disease brain lesions in high coffee consumers. ⁵⁰ Overall, beneficial effects appeared to emerge depending on the population tested and methodology and endpoint used. As described by van Dam et al., a short‐lasting arousal effect might additionally enhance cognitive performance. Although low coffee doses might result in higher alertness and shorter motor reaction times, higher exposures even reduced it. ⁵³ Thus, low‐dose stimulation results in more beneficial effects on cognitive outcomes than strong bursts of high caffeine stimulation. ⁵³ In addition, overstimulation has been reported to negatively correlate with test results ⁵³ and coffee effects have been reduced to just overcome withdrawal effects. ⁵³ , ⁵⁴ The half‐life of caffeine is on average 5 hours and may reach up to 10 hours in elderly people. ⁵⁵ , ⁵⁶ These data collectively indicate that effects on alertness may not explain the improved cognitive performance.

The FINE (Finland, Italy, and Netherlands Elderly) ¹⁵ and the CAIDE (Cardiovascular Risk Factors, Aging and Dementia) ⁴⁴ study reported an association between coffee consumption and cognitive function in a healthy population with an optimal dose of 2 to 5 units/day. Our study already found a positive association with an exposure of 1 cup/day. This is in line with recent findings, in which already 1 cup of coffee resulted in increased TMT‐B scores. ⁵⁷ Patients with AF had more comorbidities and a higher risk for cognitive impairments, which may point toward a lower response threshold.

Subgroup Analyses

Regarding the subgroup analyses, studies reported on particular positive effects among elderly ¹⁵ , ⁴¹ , ⁵¹ and female patients. ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵¹ This is in contrast to our study, in which men (with exception of the subgroup >80 years old) showed significantly higher MoCA and CoCo scores whereas women showed only a tendency. As suggested by Abbel et al, higher beneficial antioxidative effect in men might explain the difference, ⁵⁸ whereas in general elderly and more severely affected individuals with already established disease may not profit to the same extent from the benefit of coffee consumption.

Coffee Consumption and Reduced Inflammation

Our explorative association analysis aims to identify potential mechanisms for coffee consumption and cognitive function, observed a dose‐dependent reduction of the inflammatory parameters hs‐CRP and IL‐6. The reduction was observed for IL‐6 by 27% and for hs‐CRP by 22% with >5 cups/day, a robust and biologically relevant difference. The reduction of inflammatory mediators in AF patients triggered by ischemia, oxidative stress, and impaired integrity of the blood–brain‐barrier may explain our findings and is in line with recent studies. ²¹ , ²² , ²³ , ²⁵ Our findings on inflammatory markers are of 2‐fold importance because they may point at a mechanistic explanation of the effect that cannot be explained by a simple, caffeine‐induced short‐term increase of alertness in the test setting.

Potential Negative Effects

Previous reports have suggested detrimental effects of coffee on blood pressure, ⁵⁹ cholesterol levels, ⁶⁰ or induction of AF and tachyarrhythmias. ⁶¹ , ⁶² , ⁶³ More recent findings reported reduced overall cardiovascular risk with intermediate consumption (3–5 cups/day) and no additional risk for higher doses. ³⁶ , ⁶⁴ , ⁶⁵ , ⁶⁶ Two systematic reviews could not observe an association between coffee consumption and AF, even for higher doses (≥5 cups/day). ⁶⁷ They even showed a trend in the opposite direction. ⁶⁴ , ⁶⁸ Furthermore, a recent study showed no increased atrial ectopy in ambulatory, healthy caffeine consumers. ⁶⁹ Other studies reported that coffee consumption was associated with reduced risk for type 2 diabetes ⁷⁰ and an increased basal metabolic rate, ⁶⁶ suggesting additional beneficial cardiovascular effects of regular coffee consumption.

Strength and Limitations

The strengths of this study include the large sample size of more than 2400 patients, a well‐characterized cohort, detailed comorbidity evaluation, comprehensive neurocognitive test battery (including sex‐ and age‐related subanalyses), an extensive correction for confounding factors, and the novelty of the important cognitive decline in patients with AF. Furthermore, our study is one of the first to examine mechanistic anti‐inflammatory associations in patients with AF, which suggests potential protective mechanisms. The statistically significant association with higher cognitive scores appears biologically relevant and may translate into a reduction of the average “cognitive age” by 6.7 years in coffee consumers.

Limitations may consist of individual variability of habits of coffee consumption (cup size, type and strength of coffee, methods of preparation, recall bias such as misclassification of coffee intake (which in fact may represent a bias against the hypothesis), and time since the last exposure with consecutive potential short‐lasting arousal effects as discussed previously ⁶⁴ ). Interestingly, self‐reported consumption data have been shown to be surprisingly accurate and reproducible. ⁷¹ Furthermore, CYP1A2 polymorphism, ⁷² medications (eg, bronchodilators), ⁷³ and smoking ⁶⁴ may affect caffeine metabolism, but their variability might be perceived as strength since these parameters appear to be bias against the hypothesis. Further, it is in the nature of an observational study to have different subgroup size, and other potential selection biases, such as volunteer bias, may occur. Future studies may also include a premorbid cognitive function score such as the Wide Range Achievement Test, in order to illuminate cognitive changes over time.

Finally, it is important to also consider potential collider bias, particularly given our observation, that the group with the highest coffee consumption showed both the highest cardiovascular risk profile (including smoking) and the highest MoCA score. However, because a higher cardiovascular risk profile will likely associate with a lower cognitive performance, ⁷⁴ it would represent a bias against the main finding.

Conclusions

Regular coffee consumption is associated with a higher cognitive performance in patients with AF. This effect increases dose‐dependently over the range of 1 to 5 cups per day. Inflammatory markers are negatively associated and hypothesis generating (Figure 5). Further studies are required to confirm an optimal exposure of 3 to 5 cups daily in this elderly population with AF. Coffee consumption in elderly patients with AF should not be discouraged.

Figure 5. Pathophysiology.

It illustrates the strong association of increasing coffee consumption in patients with atrial fibrillation with the performance in global (eg, CoCo) and specific (eg, DSST) cognitive testing. Further, an inverse association with the inflammatory markers hs‐CRP and IL‐6 is observed. Created with Biorender.com. CoCo indicates cognitive construct; and DSST, Digit Symbol Substitution Test.

Sources of Funding

We confirm the independence of researchers from funder. The Swiss‐AF study is supported by grants from the Swiss National Science Foundation (grant numbers 33CS30_148474, 33CS30_177520, 32473B_176178, and 32003B_197524), the Swiss Heart Foundation, the Foundation for Cardiovascular Research Basel, the University of Basel and the Kardio Foundation Baden/Switzerland. All authors, external and internal, had full access to all the data and can take full responsibility for the integrity and accuracy of the data analysis.

Disclosures

None.

Supporting information

Swiss‐AF investigators Supplemental Methods

Tables S1–S8

Figures S1–S8