Habitual coffee drinkers display a distinct pattern of brain functional connectivity

The study addresses the largely unexplored chronic effects of habitual coffee (caffeine) consumption on the intrinsic functional architecture of the human brain. Caffeine acts as an adenosine A1 and A2A receptor antagonist, acutely increasing CNS excitability and affecting endurance, vigilance, dexterity, mood, memory, and cognition. While acute caffeine effects on brain activity and perfusion have been documented, the impact of long-term habitual consumption on resting-state functional connectivity remains poorly characterized. Prior work hints at potential long-term benefits (e.g., stress resistance, neuroprotection) due to receptor regulation with chronic exposure. This study aims to characterize differences in resting-state functional connectivity (static and dynamic) between habitual coffee drinkers and non-drinkers and to examine associations with psychological measures (depression, anxiety, stress) and frequency of caffeinated product consumption.

Prior fMRI studies largely focus on acute caffeine effects: increases in BOLD signal during visuomotor tasks; impacts on working memory and perfusion in older adults; increased activation in frontopolar and cingulate cortex during n-back tasks; and globally increased brain entropy, suggesting increased processing capacity. Few studies examined acute effects on resting-state functional connectivity, generally showing caffeine-induced FC reductions, links to neuro-electric power fluctuations, and stronger anticorrelations. Only one prior study touched on chronic consumption, relating coffee habits to visual cortex BOLD magnitude without assessing resting-state networks or the functional connectome. Recruiting participants with varied habitual consumption and managing abstinence present challenges, underscoring the need for comprehensive characterization using whole-brain and dynamic approaches.

Design and participants: Adults were recruited via institutional channels and media. Exclusion criteria: neurological/psychiatric disorders, habitual use of mind-altering substances, MRI contraindications. Groups: coffee drinkers (CD; ≥1 cup/day) and non-coffee drinkers (NCD; <1 cup/week). Consumption of coffee and other caffeinated products was confirmed via structured interview. All participants abstained from caffeinated products for 3 hours prior to assessment. Sample: 56 recruited (32 CD, 24 NCD); 1 CD excluded for imaging artifacts, yielding 31 CD and 24 NCD. Assessment protocol: Within ~3 hours, participants underwent: (1) interview by a certified psychologist to collect demographics; caffeine-related consumption habits; and Depression, Anxiety and Stress Scales (DASS-21). (2) MRI session. NCD completed an additional resting-state scan ~30 minutes after consuming coffee (Nespresso Ristretto, ~50 cc). Demographic/psychological analysis: Group comparisons (nonparametric Wilcoxon tests) for socio-demographics, caffeine consumption frequency, and DASS-21. Multiple regression analyses assessed associations between daily caffeinated product frequency (0 <1/day; 1 1/day; 2 2/day; 3 ≥3/day) and DASS-21, controlling for sex, age, and education. Analyses in Matlab2020a; significance p<0.05; linear regression plots in GraphPad Prism 7. MRI acquisition: Siemens Verio 3T with 32-channel head coil (Hospital de Braga). Structural T1-weighted MPRAGE: TE/TR=4.12/2420 ms, FA=9°, 1 mm isotropic voxels, FOV=176×256×256 mm³. Resting-state fMRI: multiband EPI (CMRR EPI 2D R2016A), TR/TE=1000/27 ms, FA=62°, 2 mm isotropic voxels, 64 axial slices, in-plane matrix 100×100, duration 7.5 min, eyes closed. Preprocessing: fMRIPrep 1.4.1 (Nipype 1.2.0). Full pipeline details in Supplementary Material. Resting-state analyses:

• Independent Component Analysis (ICA): Probabilistic ICA (MELODIC, FSL) identified resting-state networks (RSNs). Between-group voxelwise comparisons used FSL randomise with 5000 permutations, TFCE, family-wise error rate (FWE) controlled at α=0.05.
• Static functional connectomics: Time series from 268 regions (Shen atlas) were extracted. Pairwise Pearson correlations (Fisher r-to-z transformed) formed FC matrices. Network-Based Statistics (NBS) with 5000 permutations and FWE-corrected network significance 0.05 tested group differences.
• Dynamic functional connectivity: Leading Eigenvector Dynamics Analysis (LEIDA) assessed instantaneous phase-locking states using Shen atlas time series. Regions were matched to Yeo 7 RSNs plus cerebellar and subcortical labels for visualization. Additional analyses: Effects of acute coffee were evaluated by comparing NCD post-coffee to NCD pre-coffee (paired t-test) and to CD (independent t-test). Associations with frequency of caffeinated product consumption were assessed via multiple regression mirroring DASS-21 analyses.

Demographics: CD and NCD did not differ in age (p=0.28; Z=1.09; r=0.15) or education (p=0.07; Z=1.84; r=0.25). Frequency of caffeinated product consumption was higher in CD (p<0.001; Z=6.17; r=0.83). Sex distribution did not differ (χ²(1,N=55)=0.52, p=0.42). ICA (RSNs): Of 30 components, 15 corresponded to typical RSNs. Significant FWE-TFCE corrected differences showed higher FC in NCD (pre-coffee) relative to CD in:

• Somatosensory network with right precuneus (MNI 30, −72, 38; 7 voxels; peak t=4.4).
• Limbic network with right insula (MNI 42, −12, 2; 4 voxels; peak t=5.09). Both effects negatively correlated with frequency of caffeinated product consumption (right precuneus: p=0.003; β=−1.433; adjusted R²=0.162; right insula: p<0.001; β=−2.384; adjusted R²=0.267). After NCD consumed coffee, between-group differences diminished: somatosensory network (NCD pre vs post: t=1.86, p=0.075; NCD post vs CD: t=−2.89, p=0.006); limbic network (NCD pre vs post: t=3.88, p<0.001; NCD post vs CD: t=1.46, p=0.15). Static connectomics (NBS): One network exhibited higher connectivity in NCD (pre-coffee) than CD across thresholds 0.005–0.0005. At the most stringent threshold p=0.0005 (t-threshold=3.71, df=54; network p=0.043; Hedges’ g=1.08; 24 nodes, 46 edges), key nodes included thalamus (#262, #126), cerebellum (left anterior Culmen #245; bilateral Tonsils #238, #119), right postcentral gyrus (#33), left middle temporal gyrus (#197), left precentral gyrus (#160), bilateral caudate (#260, #122) and putamen (#124, #261). NCD post-coffee showed significantly reduced mean connectivity within this network toward CD levels (CD vs NCD post: p=0.037, t=2.13, df=54; NCD pre vs post: p=1.3×10⁻7, t=17.4, df=23). Mean NBS FC was negatively associated with caffeine consumption frequency (p<0.001; β=−0.101; adjusted R²=0.506). Dynamic FC (LEIDA): One phase-locking state (PL state 4) lasted longer in CD (17.95 ± 18.32 s) than in NCD pre-coffee (8.95 ± 6.13 s), surviving multiple-comparison correction (corrected p=0.038; medium effect size). Psychological measures: CD displayed higher stress levels than NCD. While no other group effects were observed in psychological assessment, higher frequency of caffeine consumption was associated with increased anxiety in males. Overall: Habitual coffee drinkers exhibited decreased resting-state FC in somatosensory and limbic networks, reduced connectivity within a subcortical/posterior network related to somatosensory, motor, and emotional processing, and prolonged occupancy of a subcortical–visual–cerebellar dynamic state. These effects scaled with frequency of caffeinated product consumption and were reproducible acutely after coffee intake in NCD.

The findings demonstrate that habitual coffee consumption is associated with distinct alterations in resting-state functional connectivity. Reduced connectivity within somatosensory and limbic networks, as well as within a broader subcortical/cerebellar/posterior cortical network, suggests that chronic caffeine exposure modulates circuits implicated in sensory processing, motor readiness, and emotional regulation. Dynamic FC results further indicate longer dwell time in a subcortical–visual–cerebellar phase-locking state among habitual consumers, aligning with potential effects on alertness and readiness to action. Crucially, the dependence of these effects on consumption frequency and their acute reproduction in non-drinkers after coffee intake support a direct link between caffeine consumption and the observed network patterns. Psychologically, higher stress in CD and a positive association between consumption frequency and anxiety in males provide behavioral correlates that may relate to altered FC within limbic and arousal-related circuits. Collectively, the results extend prior acute caffeine literature by characterizing chronic consumption-related differences in both static and dynamic intrinsic brain networks.

Higher consumption of coffee and caffeinated products impacts brain functional connectivity at rest, characterized by reduced connectivity in somatosensory and limbic networks, altered subcortical/posterior network connectivity, and longer occupancy of a subcortical–visual–cerebellar dynamic state, with implications for emotionality, alertness, and readiness to action. These effects scale with consumption frequency and are reproducible after acute coffee intake in non-drinkers.