research paper of coffee

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Open access
• Published: 20 April 2021

Habitual coffee drinkers display a distinct pattern of brain functional connectivity

• Ricardo Magalhães 1 , 2 , 3 , 4   na1 ,
• Maria Picó-Pérez   ORCID: orcid.org/0000-0002-1573-2445 1 , 2 , 3   na1 ,
• Madalena Esteves 1 , 2 , 3   na1 ,
• Rita Vieira   ORCID: orcid.org/0000-0001-6762-406X 1 , 2 , 3 ,
• Teresa C. Castanho 1 , 2 , 3 ,
• Liliana Amorim 1 , 2 , 3 ,
• Mafalda Sousa 1 , 2 , 3 ,
• Ana Coelho 1 , 2 , 3 ,
• Henrique M. Fernandes 5 ,
• Joana Cabral   ORCID: orcid.org/0000-0002-6715-0826 1 , 2 , 3 , 5 ,
• Pedro S. Moreira 1 , 2 , 3 , 6 &
• Nuno Sousa   ORCID: orcid.org/0000-0002-8755-5126 1 , 2 , 3 , 7  

Molecular Psychiatry volume  26 ,  pages 6589–6598 ( 2021 ) Cite this article

143k Accesses

32 Citations

489 Altmetric

Metrics details

• Neuroscience

Coffee is the most widely consumed source of caffeine worldwide, partly due to the psychoactive effects of this methylxanthine. Interestingly, the effects of its chronic consumption on the brain’s intrinsic functional networks are still largely unknown. This study provides the first extended characterization of the effects of chronic coffee consumption on human brain networks. Subjects were recruited and divided into two groups: habitual coffee drinkers (CD) and non-coffee drinkers (NCD). Resting-state functional magnetic resonance imaging (fMRI) was acquired in these volunteers who were also assessed regarding stress, anxiety, and depression scores. In the neuroimaging evaluation, the CD group showed decreased functional connectivity in the somatosensory and limbic networks during resting state as assessed with independent component analysis. The CD group also showed decreased functional connectivity in a network comprising subcortical and posterior brain regions associated with somatosensory, motor, and emotional processing as assessed with network-based statistics; moreover, CD displayed longer lifetime of a functional network involving subcortical regions, the visual network and the cerebellum. Importantly, all these differences were dependent on the frequency of caffeine consumption, and were reproduced after NCD drank coffee. CD showed higher stress levels than NCD, and although no other group effects were observed in this psychological assessment, increased frequency of caffeine consumption was also associated with increased anxiety in males. In conclusion, higher consumption of coffee and caffeinated products has an impact in brain functional connectivity at rest with implications in emotionality, alertness, and readiness to action.

Similar content being viewed by others

Drinking coffee enhances neurocognitive function by reorganizing brain functional connectivity

Resting-state functional connectivity and structural differences between smokers and healthy non-smokers

Brain activity during a working memory task after daily caffeine intake and caffeine withdrawal: a randomized double-blind placebo-controlled trial

Introduction.

Coffee is the most widely consumed beverage, with particular interest for human health in view of its short-term effects on attention, sleep, and memory and its long-term impact on the appearance of different diseases and on healthy span of ageing [ 1 , 2 ]. Coffee has several constituents able to impact on human health, amongst which stems caffeine, which is the most widely consumed psychostimulant in the world [ 3 ]. Despite its widespread use it is surprising to note that a thorough characterization of the chronic effects of coffee upon the human brain is still lacking. In the present work we aim to begin addressing that issue.

In the brain, caffeine acts as an antagonist of adenosine A1 and A2A receptors, leading to hyperexcitability of the central nervous system [ 3 , 4 ]. This induces acute effects in diverse domains, such as physical endurance [ 1 , 5 ], vigilance, dexterity [ 6 ], mood [ 7 , 8 ], memory [ 9 ], and cognitive function [ 1 , 8 , 10 ]. There is also evidence that coffee/caffeine intake can normalize anxiety [ 11 ], although higher doses of caffeine may be anxiogenic [ 1 , 12 ] by disrupting the HPA axis [ 13 ]. On the other hand, epidemiological and animal studies converge in concluding that coffee, caffeine and adenosine receptor antagonists attenuate the burden of neurodegenerative disorders such as Alzheimer’s [ 14 ], or psychiatric disorders such as depression [ 15 ]. Indeed, chronic antagonism of either A1 or A2 receptors seems to induce an upregulation of the former, but not the latter. The resulting altered receptor ratio may explain the shift from the acute psychomotor effects (e.g., attention, vigilance) to the longer-term actions of coffee (e.g., stress resistance, neuroprotection) effects [ 4 , 16 ].

Functional magnetic resonance imaging (fMRI) allows studying, in a noninvasive way, the function of the human brain during execution of different tasks or at rest [ 17 ]. So far, most studies using fMRI were focused on measuring the acute effects of caffeine intake in the brain. Briefly, they have reported caffeine-related increases in blood oxygenation-dependent-level (BOLD) signal in different cortical and subcortical areas during a visuomotor task [ 18 ]; an impact in working memory and perfusion in elderly subjects [ 19 , 20 ]; an increase in BOLD activation in the frontopolar and cingulate cortex during a 2-back verbal working memory task [ 21 , 22 ]; and a global caffeine-induced increase in brain entropy, possibly representing an increased processing capacity [ 23 ]. Very few studies, however, were performed to study the acute effects of caffeine in functional connectivity (FC) at rest [ 24 , 25 ]. Those few studies reveal a general trend for a caffeine-induced reduction in FC, associated with neuro-electric power fluctuations as measured through magnetoencephalography and exacerbated anticorrelations. Despite this existing literature, many aspects of the characterization of the impact of caffeine on the brain remain unexplored. Critical amongst these is the characterization of the chronic effects of habitual coffee and caffeine consumption upon the functional architecture of the brain. We are only aware of a single study that touched on this subject [ 26 ]. That work revealed an association between different habits of coffee consumption and the magnitude of BOLD signals in the visual cortex; however, it did not address possible effects on the functional connectome or resting state networks. Pursuit of the latter can present significant challenges in finding and recruiting participants with sufficient variation in consumption habits and who are willing to undergo necessary, even if short, abstinence procedures.

To tackle this gap, herein we will use whole brain approaches [ 27 , 28 , 29 ], as well as the study of brain functional dynamics [ 30 ] to compare FC and its dynamics between habitual and non-habitual coffee consumers. In addition, and because of the potential anxiogenic and HPA-disrupting role of caffeine, measures of psychological state (depression, anxiety, and stress) will also be acquired, in order to explore the potential association of habitual coffee consumption with these variables.

Subject recruitment and assessment

Participants were recruited through advertisement on the Institute’s social media, institutional e-mail, and press releases distributed among Portuguese local and national newspapers. Exclusion criteria included the presence of neurological or psychiatric disorders, habitual consumption of mind-altering substances, and the inability to undergo MRI. Two experimental groups were created according to participants’ coffee consumption habits: coffee drinkers (CD), who drank a minimum of one cup of caffeinated coffee per day; and non-coffee drinkers (NCD), who had no habits of regular consumption of coffee (less than one cup per week). Consumption of coffee as well as other caffeinated products was confirmed in a structured interview. Participants were instructed to abstain from caffeinated products for 3 h before the assessment, in order to avoid acute influence of caffeine. Fifty-six subjects were recruited (32 CD and 24 NCD). One participant from the CD group was excluded due to imaging artifacts, rendering a final sample of 31 CD and 24 NCD. Characterization of subjects was done in two (CD) or three (NCD) parts within a 3 h time-period: participants were first interviewed by a certified psychologist. This was followed by an MRI scanning session, and, in the case of the NCD, the first scanning session was followed by ingestion of coffee (Nespresso ® Ristretto, ~50 cc) before a rs-fMRI scan ~30 min thereafter. During the interview, the following data were gathered: demographic data; habits of coffee and other caffeinated products consumption; and assessment of depression, anxiety, and stress scores through the Depression, Anxiety and Stress Scales (DASS-21, [ 31 , 32 ]).

Demographic and psychological data analysis

CD and NCD groups were compared in terms of sociodemographic variables, frequency of consumption of caffeinated products, and psychological variables. Since the variables did not follow a normal distribution, nonparametric tests were applied (Wilcoxon test). Moreover, multiple regression analyses were performed, aiming to determine the association between daily consumption of caffeinated products such as coffee, tea, chocolate, etc. (0 = <1/day; 1 = 1/day; 2 = 2/day; 3 = 3 or more/day) and the psychological data measured with the DASS-21 questionnaire (controlled for sex, age, and education), independently of the groups. These analyses were performed on Matlab2020a software (The Mathworks, Inc.) and p  < 0.05 was considered the threshold for statistical significance. Linear regression representations were generated in Prism 7 software (GraphPad Software, Inc.).

MRI brain imaging

Magnetic resonance imaging scans were conducted using a Siemens Verio 3T (Siemens, Erlangen, Germany) located in Hospital de Braga (Braga, Portugal) using a 32-channel head antenna. The scanning session included as an anatomical acquisition a T1-weighted sagittal magnetization-prepared rapid acquisition with gradient echo (TE/TR = 2420/4.12 ms, FA = 9°, 1 mm 3 isometric voxel size, Field-of-View = 176 × 256 × 256 mm 3 ). The resting-state fMRI (rs-fMRI) acquisition used a multi-band echo planar imaging sequence, CMRR EPI 2D (R2016A, Center for Magnetic Resonance Research, University of Minnesota, Minnesota, USA [ 33 , 34 , 35 ]) sensitive to fluctuations in the BOLD contrast (TR/TE = 1000/27 ms, FA = 62°, 2 mm 3 isometric voxel size, 64 axial slices over an in plane matrix of 100 × 100). The rs-fMRI acquisition had a duration of 7.5 min, during which subjects were instructed to remain with their eyes closed, relaxed, and let their minds wander freely.

Preprocessing of MRI data

MRI results included in this manuscript were preprocessed using fMRIPrep 1.4.1 ([ 36 ]; RRID:SCR_016216), which is based on Nipype 1.2.0 ([ 37 , 38 ]; RRID:SCR_002502). A full description of the preprocessing pipeline can be found in the Supplementary material.

Resting-state analysis

Independent component analysis.

Resting-state network (RSN) maps were analyzed voxel-wise through a probabilistic independent component analysis (ICA) as implemented in Multivariate Exploratory Linear Optimized Decomposition into Independent Components, distributed with FSL [ 39 ]. For further details check the Supplementary material.

The RSNs FC was compared between CD and NCD groups, using a nonparametric permutation procedure implemented in the randomize tool from FSL [ 40 ]. Threshold-free cluster enhancement (TFCE) was used to detect widespread significant differences and control the family-wise error rate (FWE-R) at α  = 0.05. In total, 5000 permutations were performed.

Static functional connectomics analysis

To assess differences between the two groups in the functional connectome, the mean time series of the 268 regions of the Shen Atlas [ 41 ] were extracted. The Pearson correlation between time series, followed by Fisher r-to-Z transformation, were calculated to obtain matrices of FC for each subject. To overcome the issue of multiple comparisons induced by the large number of connections in the network, we applied the network-based statistics (NBS) approach [ 42 ]. A total of 5000 permutations were used, together with a FWE corrected network significance of 0.05. For more details check the Supplementary material.

Dynamic functional analysis

We applied the leading eigenvector dynamics analysis (LEiDA, [ 30 ]) approach to study the changes in the functional dynamics associated with habitual caffeine consumption. Instantaneous FC was calculated for each subject at each time point for all 268 regions of interest of the Shen atlas, using the time series extracted for the static analysis. To help visually identify phase locked (PL) states, the overlap between each anatomical region of each state to the 7 Yeo RSN’s [ 43 ], plus two other labels for the cerebellum and subcortical units, was calculated and anatomical units color coded in accordance to the best match. For more details check the reference paper or the Supplementary material.

Effects of acute coffee consumption and frequency of caffeine consumption

The significant findings obtained with ICA, NBS, and LEiDA were further explored, aiming to assess the effects of acute coffee consumption in NCD and of frequency of consumption of caffeinated products in both groups. The first were assessed by comparing NCD after coffee consumption with data from CD (independent sample t -test) and NCD (before coffee consumption; paired sample t -test). The second were evaluated by performing multiple regression analyses following the same approach described for the DASS-21 questionnaire.

Demographic analysis

CD and NCD groups did not differ in terms of age (range 19–57; p  = 0.28; Z  = 1.09; r  = 0.15) or number of formal years of education (range 12–25; p  = 0.07; Z  = 1.84; r  = 0.25). Frequency of consumption of caffeinated products was, as expected, higher in the CD group ( p  < 0.001; Z  = 6.17; r  = 0.83). Sex distribution was not significantly different between groups ( χ 2 (1, N  = 55) = 0.52, p  = 0.42), despite the CD group presenting a slightly higher proportion of males (41.94%) in comparison with the NCD group (33.33%). Descriptive statistics can be found in Table  1 .

Effect of habitual caffeine consumption on rs-fMRI data

Independent components analysis.

Thirty components were obtained from the probabilistic ICA of CD and NCD (before consuming coffee). Fifteen of these components were found to be representative of the most typical RSNs. A tendency toward lower FC patterns in the CD group can be seen in most of these networks (see Supplementary Fig.  1 ). Despite this, we only found significant FWE-R TFCE corrected between-group differences in two of them, namely, in the somatosensory network and the limbic network (Fig.  1 ). Regarding the somatosensory network, NCD presented a pattern of higher connectivity with the right precuneus (MNI coordinates = 30, −72, 38; 7 voxels; peak t value = 4.4). Moreover, for the limbic network, NCD had higher FC in the right insula compared to CD (MNI coordinates = 42, −12, 2; 4 voxels; peak t value = 5.09). Of note, these effects were also linearly associated with the caffeinated products’ frequency of consumption. Negative correlations were found for both right precuneus ( p  = 0.003; β  = −1.433; adjusted R 2  = 0.162; Fig.  1B ) and right insula ( p  < 0.001; β  = −2.384; adjusted R 2  = 0.267; Fig.  1B ). Detailed statistics can be found in Supplementary Table  1 .

A Sagittal, coronal, and axial view of the clusters showing significant between-group differences in the connectivity between the somatosensory network and the right precuneus (top) and the limbic network and the right insula (bottom). The FWE-R TFCE corrected clusters are shown in dark blue overlaid over a more extended non-significant after multiple comparison correction cluster in hot color scale scheme, for visualization purposes. B Associations of frequency of consumption of caffeinated beverages with the mean FC of the right precuneus and the right insula. C Scatter plots showing the mean FC of the right precuneus and the right insula for the NCD before drinking coffee (NCD), the NCD after drinking coffee (NCD pos), and the CD.

Importantly, the group differences described were reduced after NCD drank coffee (see Fig.  1C ; somatosensory network: pre vs post NCD t value = 1.86, p  = 0.075, post NCD vs CD t value = −2.89, p  = 0.006; limbic network: pre vs post NCD t value = 3.88, p  < 0.001, post NCD vs CD t value = −1.46, p  = 0.15). This points to a potential causality link between coffee drinking and the above-described changes in lower connectivity in the somatosensory and in the limbic networks.

Connectomics analysis

From the connectomics analysis done using NBS, a single network of significantly higher connectivity was found in the NCD group (pre-coffee) between the thresholds of 0.005 and 0.0005 (for statistics of all thresholds see Supplementary Table  2 ). For ease of visualization, we present only the results found at the highest significant threshold of p  = 0.0005 ( t (threshold) = 3.71, df = 54, p (network) = 0.043, Hedge’s g  = 1.08 (large effect size), 24 nodes, 46 edges; Fig.  2A ). The full list of nodes with significant different edges between groups across all thresholds can be found in Supplementary Table  3 . Of these we highlight the Thalamus (nodes #262 and #126), the Cerebellum (left anterior Culmen #245 and bilateral Tonsils #238 and #119), the right Postcentral Gyrus (#33), the left Middle Temporal Gyrus (#197), the left Precentral Gyrus (#160), and the bilateral Caudate (#260 and #122) and Putamen (#124 and #261) as having the most strongly affected connections within the network.

A Sagittal, coronal, and axial view of the network with nodes and edges colored in red–yellow color scheme representing the statistical t value of the difference between groups. B Scatter plot of the mean FC within the significant network for each experimental group. C Associations of frequency of consumption of caffeinated beverages with the mean FC of the network found in NBS.

When observing the average network connectivity from this network, NCD post-coffee drink displayed a significant reduction in connectivity (Fig.  2B ), leading to a profile more similar to CD ( p  = 0.037, t  = 2.13, df = 54) than to NCD pre-coffee drink ( p  = 1.3 × 10 −7 , t  = 7.4, df = 23). NBS mean FC was negatively associated with frequency of caffeine consumption ( p  < 0.001; β  = −0.101; adjusted R 2  = 0.506; Fig.  2C ). Detailed statistics can be found in Supplementary Table  1 .

From the dynamic FC analysis, one functional subsystem (Fig.  3A , PL state 4) was found to last significantly longer in CD (Fig.  3B , 17.95 ± 18.32 s) compared to pre-coffee NCD (8.95 ± 6.13 s) surviving correction for multiple comparisons with a corrected p  = 0.038 and a medium effect size with Hedge’s g  = 0.62. No BOLD phase-locking state was found to significantly differ in terms of probability of occurrence (see Supplementary Table  4 for all p values for all partition models).

A sagittal and axial views representing the state anatomical areas of each phase locked (PL) state. B Bar plot representing the group differences between coffee and non-coffee drinkers. Differences of p  < 0.05 are indicated in red, while multiple comparison surviving effects are indicated in green. C Associations of frequency of consumption of caffeinated beverages with the average duration (in seconds) of PL state 4. D Bar plot of the probability of state 4 for the CD, NCD, and NCD post caffeine consumption groups. E Life time of state 4 for the CD, NCD, and NCD post caffeine consumption groups. F Colored labels used to match each anatomical area of the PL states to different resting state networks.

This BOLD phase-locking state, corresponding to the fourth most probable state when partitioning the data into nine states, comprises a large number of nodes in the cerebellum, visual network as well as several subcortical nodes such as the bilateral thalamus and parahippocampal gyrus (mapped and color coded through the reference shown in Fig.  3F ). While this was the only result that survived correction for multiple comparisons, it is relevant to note that the equivalent LEiDA state for k  = 10 is just below the threshold ( p  = 0.051, Supplementary Table  4 and Supplementary Figs.  2 and 3 ). Furthermore, LEiDA lifetime results were positively correlated with frequency of caffeine consumption ( p  = 0.012; β  = 2.176; adjusted R 2  = 0.083; Fig.  3C ).

After drinking coffee, both the lifetime and the probability of this state in NCD became closer to the values observed in CD, with the probability not being significantly different from CD ( p  = 0.5, t  = 0.67, df = 54), while being significantly higher than NCD pre-coffee ( p  = 0.037, t  = 2.31, df = 23, Fig.  3D ). For the life time of state 4, post-coffee drink NCD were not significantly different from CD ( p  = 0.177, t  = 1.37, df = 54) nor the pre-drink NCD ( p  = 0.107, t  = 1.68, df = 23, Fig.  3E ). All results across the different k’ s can be found in Supplementary Figs.  2 and 3 and Supplementary Table  4 .

Effect of habitual caffeine consumption on psychological data

The association between coffee consumption and stress, anxiety, and depression (DASS-21) was assessed. When comparing CD and NCD groups, only stress was significantly different between groups (stress— p  = 0.025; Z  = 2.237; r  = 0.307; anxiety— p  = 0.851; Z  = −0.188; r  = −0.026; depression— p  = 0.085; Z  = 1.724; r  = 0.237), with CD showing higher levels of stress than NCD (median (Med) = 6.0; interquartile range (IQR) = 6.0 vs Med = 4.0; IQR = 4.0, respectively). Of notice, particular items of the DASS-21 Stress subscale that can be related to arousal were increased in CD. Items #1 and #12, which measure difficulty to relax, presented statistically significant differences ( p  = 0.007, Mann–Whitney test), while item #8, that relates to nervous arousal, presented a trend in the same direction ( p  = 0.083). Interestingly, item #7 (Anxiety subscale), that is associated with skeletal musculature, despite not achieving a statistically significant difference between groups, tended to be lower in CD ( p  = 0.113), suggesting a segregation between the motor and arousal loops.

When assessing the effects of frequency of caffeine consumption in self-reported variables (controlling for sex, age, and education), the positive correlation with stress was maintained ( p  = 0.004; β  = 1.292; adjusted R 2  = 0.135; Fig.  4A ). Moreover, a sex by anxiety interaction was found ( p  = 0.023; β  = 0.683; adjusted R 2  = 0.085; Fig.  4B ), which seems to be driven by a positive correlation in males. No significant effects were found for the depression subscale ( p  = 0.128; β  = 0.450; adjusted R 2  = 0.108; Fig.  4C ). Detailed statistics can be found in Supplementary Table  1 .

Associations of frequency of consumption of caffeinated products with the DASS-21 subscales of stress ( A ) and anxiety ( B ), and non-significant association with the depression subscale ( C ).

Herein we describe for the first time the effects of habitual coffee consumption on the human brain networks. We show that habitual CD have different patterns of FC in comparison with NCD. Our rs-fMRI analysis revealed decreased FC of the somatosensory and limbic networks in CD that correlated with the frequency of consumption of caffeinated products. Such changes were replicated in NCD after a single coffee, suggesting possible causality between coffee intake and altered patterns of brain network connectivity. Previous studies have described a reduction of similar RSN connectivity after acute caffeine ingestion [ 25 , 44 ].

Decreased FC in somatosensory and related networks in CD likely represents a more efficient and beneficial pattern of connections with respect to motor control and alertness; importantly this fits our findings of trends of increased scores in CD in the specific items of the DASS-21 scale that measure these dimensions. The other network impacted by coffee intake was the limbic network, which is involved in processing the sensory input from the external and internal environment which, by modulating memory and motivation, determine emotional, autonomic, motor, and cognitive responses [ 45 ]. A previous resting-state PET study showed reduced metabolic activity in components of this network after caffeine ingestion [ 18 ] and a study using a hedonic fMRI task showed decreased activation in neuronal areas associated with memory and reward [ 46 ] in caffeine consumers compared to non-consumers; the present FC data are consistent with those reports.

Analysis of the global functional connectome using NBS revealed a network impacted by the habitual consumption of caffeine. This widespread network of reduced FC comprised cerebellar, subcortical (striatal and thalamic), and motor cortex regions, partially matching previously reported effects of acute caffeine ingestion [ 24 , 25 ]. Interestingly, there is a clear bilateral involvement of striatal nodes and of the thalamus which, respectively, have the highest densities of A2A and A1 receptors in the brain [ 47 , 48 ]. The action of caffeine in these regions has an influence on cortico-striatal-thalamic and cerebellar-thalamocortical loops that are relevant for a variety of neuronal processes. Thus, the observed decrease in FC at rest in this network in regular caffeine-ingesting individuals reveals greater segregation of these areas with less inter-regional dependencies, favoring greater efficiency within these loops. It is relevant to note here that, even though A1 and A2A receptors are thought to mediate differential actions [ 49 ], similar effects were observed in both loops. This likely reflects the fact that fMRI provides proxy aggregate measurements of functional connections among a network of brain areas.

A previous study reported that caffeine increases brain entropy, indicating higher information processing capacity across the cerebral cortex [ 23 ]. Our LEiDA analysis revealed a dynamic state involving several cerebellar and subcortical areas, with a longer average lifetime in habitual CD. This network comprises several nodes, including the cerebellum, thalamus, and parahipocampal, lingual, and inferior occipital gyri which are relevant in the context of caffeine consumption—caffeine is known to decrease mind wandering [ 50 ] and to increase attention, alertness, and arousal [ 51 ]. In fact, the nodes implicated in this network are linked by visual processing; among these, the thalamus is critical for distributing cognitive control [ 52 ]. The lingual and inferior occipital gyrus are also implicated in visual processing, while the parahippocampus is involved in memory encoding and retrieval [ 20 , 21 ]; the latter may explain why caffeine reportedly facilitates memory processes [ 9 ]. Lastly, evidence of strong rsFC between the cerebellum, known to be also implicated in sensory processing [ 53 ] and a number of sensorial cortices [ 54 ], explains the observed increased visual alertness/attention and readiness to sensorial perception among CD individuals. While similar findings have been previously reported [ 6 ], only one other study assessed habitual CD using MRI, and did not characterize changes in FC [ 26 ]. Importantly, similarly to the other neuroimaging findings, a common pattern of connectivity dynamics was found in CD individuals and NCD subjects who drank a single coffee before scanning.

In order to provide a link with other neuropsychologic dimensions, we also assessed our subjects in the DASS-21. Interestingly, we observed habitual CD to display increased levels of stress; there was a clear positive association between the indices of stress and the amount of consumption of caffeinated drinks. Interestingly, items of the DASS-21 sub-score that showed greater variance between CD vs NCD were those related with difficulty to relax (items #1 and #12), and those related to nervous arousal (item #8), consistent with the common attribution of alertness and arousal to coffee intake. It also deserves to be mentioned that, despite the display of a higher anxiety among CD (particularly in males), there was a decrease in DASS-21 item (#7) which matches the effects on the skeletal muscles in CD; this, in turn, fits the findings of better segregation of the above-described loops. The present results extend previous studies that described an association between coffee/caffeine consumption and stress and anxiety [ 1 , 13 , 16 , 55 ] and sex [ 13 , 16 ]. It is important to note, however, that causality cannot be inferred from our study design. Our results are open to two interpretations: higher coffee/caffeine consumption leads to increased stress and anxiety; or, alternatively, higher stress and anxiety induce higher coffee/caffeine consumption. Moreover, given that resting-state studies using stress and anxiety samples have shown both decreases and increases in FC [ 56 , 57 , 58 ], the possibility that coffee/caffeine consumption elicits decreases in FC or compensates for FC beyond a certain threshold, must also be considered. While the first possibility is in line with studies showing increased anxiety upon both acute caffeine administration in humans [ 1 , 12 ] and prolonged ingestion in rodents [ 59 ] reports that greater caffeine consumption under periods of stress may help maintain synaptic homeostasis [ 60 ] as well as prevent mood disorders warrant further study in future.

The methodologies applied in the present study do not allow us to draw precise relationships between the psychological and neuroimaging results and the dosage and metabolism of caffeine among individual subjects. To study the individual responses to the acute and chronic effects of caffeinated product intake would be a complex undertaking, requiring subjects to adapt their daily habits and strict abstinence regimens. Based on our experience, recruitment of subjects for a properly balanced study is also difficult since NCD subjects are insufficiently motivated to engage in studies on the actions of caffeine. Nevertheless, we are currently developing alternative strategies that would allow us to deliver calibrated doses of caffeine during fMRI scanning sessions to better discriminate its effects from other factors (e.g., stress). Our future work will also examine inter-individual differences in response to caffeine consumption, the subjective experience of coffee consumption, as well as the influence of additional factors as the consumption of alcohol and tobacco. Despite such gaps, the data presented here represent a contribution to the knowledge of the “caffeinated brain” and how these changes underlie the behavioral effects triggered by coffee intake, with implications for physiological and pathological conditions.

Code availability

In-house scripts used in the NBS analysis are fully available online at open science framework website ( https://osf.io/qepc8/ ) and LEiDA scripts at github ( https://github.com/juanitacabral/LEiDA ).

McLellan TM, Caldwell JA, Lieberman HR. A review of caffeine’s effects on cognitive, physical and occupational performance. Neurosci Biobehav Rev. 2016;71:294–312.

Article   CAS   PubMed   Google Scholar  

O’Keefe JH, DiNicolantonio JJ, Lavie CJ. Coffee for cardioprotection and longevity. Prog Cardiovascular Dis. 2018;61:38–42.

Article   Google Scholar  

Fredholm BB, Bättig K, Holmén J, Nehlig A, Zvartau EE. Actions of caffeine in the brain with special reference to factors that contribute to its widespread use. Pharmacol Rev. 1999;51:83–133.

CAS   PubMed   Google Scholar  

Ribeiro JA, Sebastião AM. Caffeine and Adenosine. J Alzheimer’s Dis. 2010;20:S3–15.

Article   CAS   Google Scholar  

Southward K, Rutherfurd-Markwick KJ, Ali A. The effect of acute caffeine ingestion on endurance performance: a systematic review and meta–analysis. Sports Med. 2018;48:1913–28.

Article   PubMed   Google Scholar  

Killgore WDS, Kamimori GH. Multiple caffeine doses maintain vigilance, attention, complex motor sequence expression, and manual dexterity during 77 h of total sleep deprivation. Neurobiol Sleep Circadian Rhythms. 2020;9:100051.

Article   PubMed   PubMed Central   Google Scholar  

Sane RM, Jadhav PR, Subhedar SN. The acute effects of decaffeinated versus caffeinated coffee on reaction time, mood and skeletal muscle strength. J Basic Clin Physiol Pharmacol. 2019;30.

Smit HJ, Rogers PJ. Effects of low doses of caffeine on cognitive performance, mood and thirst in low and higher caffeine consumers. Psychopharmacology (Berl). 2000;152:167–73.

Borota D, Murray E, Keceli G, Chang A, Watabe JM, Ly M, et al. Post-study caffeine administration enhances memory consolidation in humans. Nat Neurosci. 2014;17:201–3.

Article   CAS   PubMed   PubMed Central   Google Scholar  

Franceschini S, Lulli M, Bertoni S, Gori S, Angrilli A, Mancarella M, et al. Caffeine improves text reading and global perception. J Psychopharmacol. 2020;34:315–25.

Harris A, Ursin H, Murison R, Eriksen HR. Coffee, stress and cortisol in nursing staff. Psychoneuroendocrinology 2007;32:322–30.

Bruce M. Anxiogenic effects of caffeine in patients with anxiety disorders. Arch Gen Psychiatry. 1992;49:867.

O’Neill CE, Newsom RJ, Stafford J, Scott T, Archuleta S, Levis SC, et al. Adolescent caffeine consumption increases adulthood anxiety-related behavior and modifies neuroendocrine signaling. Psychoneuroendocrinology. 2016;67:40–50.

Eskelinen MH, Kivipelto M. Caffeine as a protective factor in dementia and Alzheimer’s disease. J Alzheimer’s Dis. 2010;20:S167–74.

Wang L, Shen X, Wu Y, Zhang D. Coffee and caffeine consumption and depression: a meta-analysis of observational studies. Aust NZ J Psychiatry. 2016;50:228–42.

Kaster MP, Machado NJ, Silva HB, Nunes A, Ardais AP, Santana M, et al. Caffeine acts through neuronal adenosine A2A receptors to prevent mood and memory dysfunction triggered by chronic stress. Proc Natl Acad Sci USA. 2015;112:7833–8.

Soares JM, Magalhães R, Moreira PS, Sousa A, Ganz E, Sampaio A, et al. A Hitchhiker’s guide to functional magnetic resonance imaging. Front Neurosci. 2016;10:515.

Park CA, Kang CK, Son YD, Choi EJ, Kim SH, Oh ST, et al. The effects of caffeine ingestion on cortical areas: functional imaging study. Magn Reson Imaging. 2014;32:366–71.

Haller S, Rodriguez C, Moser D, Toma S, Hofmeister J, Sinanaj I, et al. Acute caffeine administration impact on working memory-related brain activation and functional connectivity in the elderly: a BOLD and perfusion MRI study. Neuroscience. 2013;250:364–71.

Klaassen EB, de Groot RHM, Evers EAT, Snel J, Veerman ECI, Ligtenberg AJM, et al. The effect of caffeine on working memory load-related brain activation in middle-aged males. Neuropharmacology. 2013;64:160–7.

Haller S, Montandon ML, Rodriguez C, Moser D, Toma S, Hofmeister J, et al. Caffeine impact on working memory-related network activation patterns in early stages of cognitive decline. Neuroradiology. 2017;59:387–95.

Koppelstaetter F, Poeppel TD, Siedentopf CM, Ischebeck A, Verius M, Haala I, et al. Does caffeine modulate verbal working memory processes? An fMRI study. NeuroImage. 2008;39:492–9.

Chang D, Song D, Zhang J, Shang Y, Ge Q, Wang Z. Caffeine caused a widespread increase of resting brain entropy. Sci Rep. 2018;8:2700.

Tal O, Diwakar M, Wong CW, Olafsson V, Lee R, Huang MX, et al. Caffeine-induced global reductions in resting-state BOLD connectivity reflect widespread decreases in MEG connectivity. Front Hum Neurosci. 2013;7:63.

Wong CW, Olafsson V, Tal O, Liu TT. Anti-correlated networks, global signal regression, and the effects of caffeine in resting-state functional MRI. NeuroImage. 2012;63:356–64.

Laurienti PJ, Field AS, Burdette JH, Maldjian JA, Yen Y-F, Moody DM. Dietary caffeine consumption modulates fMRI measures. NeuroImage. 2002;17:751–7.

Bullmore E, Sporns O. Complex brain networks: graph theoretical analysis of structural and functional systems. Nat Rev Neurosci. 2009;10:186–98.

Deco G, Kringelbach ML. Great expectations: using whole-brain computational connectomics for understanding neuropsychiatric disorders. Neuron. 2014;84:892–905.

Sporns O, Chialvo D, Kaiser M, Hilgetag C. Organization, development and function of complex brain networks. Trends Cogn Sci. 2004;8:418–25.

Cabral J, Vidaurre D, Marques P, Magalhães R, Silva Moreira P, Miguel, et al. Cognitive performance in healthy older adults relates to spontaneous switching between states of functional connectivity during rest. Sci Rep. 2017;7:5135.

Lovibond PF, Lovibond SH. The structure of negative emotional states: comparison of the Depression Anxiety Stress Scales (DASS) with the beck depression and anxiety inventories. Behav Res Ther. 1995;33:335–43.

Pais-Ribeiro JL, Honrado A, Leal I. Contribuição para o Estudo da Adaptação Portuguesa das Escalas de Ansiedade, Depressão e Stress (EADS) de 21 itens de Lovibond e Lovibond. Psic Saúde Doenç. 2004;5:229–39.

Google Scholar  

Feinberg DA, Moeller S, Smith SM, Auerbach E, Ramanna S, Glasser MF, et al. Multiplexed echo planar imaging for sub-second whole brain FMRI and fast diffusion imaging. PLoS ONE. 2010;5:e15710.

Moeller S, Yacoub E, Olman CA, Auerbach E, Strupp J, Harel N, et al. Multiband multislice GE-EPI at 7 tesla, with 16-fold acceleration using partial parallel imaging with application to high spatial and temporal whole-brain fMRI. Magn Reson Med. 2010;63:1144–53.

Xu J, Moeller S, Auerbach EJ, Strupp J, Smith SM, Feinberg DA, et al. Evaluation of slice accelerations using multiband echo planar imaging at 3T. NeuroImage. 2013;83:991–1001.

Esteban O, Markiewicz CJ, Blair RW, Moodie CA, Isik AI, Erramuzpe A, et al. fMRIPrep: a robust preprocessing pipeline for functional MRI. Nat Methods. 2019;16:111–6.

Gorgolewski KJ, Esteban O, Ellis DG, Notter MP, Ziegler E, Johnson H, et al. Nipype: a flexible, lightweight and extensible neuroimaging data processing framework in Python. 0.13.1. Zenodo. 2017.

Gorgolewski K, Burns CD, Madison C, Clark D, Halchenko YO, Waskom ML, et al. Nipype: a flexible, lightweight and extensible neuroimaging data processing framework in python. Front Neuroinform. 2011;5:13.

Beckmann CF, Smith SM. Probabilistic independent component analysis for functional magnetic resonance imaging. IEEE Trans Med Imaging. 2004;23:137–52.

Winkler AM, Ridgway GR, Webster MA, Smith SM, Nichols TE. Permutation inference for the general linear model. NeuroImage. 2014;92:381–97.

Shen X, Tokoglu F, Papademetris X, Constable RT. Groupwise whole-brain parcellation from resting-state fMRI data for network node identification. NeuroImage. 2013;82:403–15.

Zalesky A, Fornito A, Bullmore ET. Network-based statistic: identifying differences in brain networks. NeuroImage. 2010;53:1197–207.

Thomas Yeo BT, Krienen FM, Sepulcre J, Sabuncu MR, Lashkari D, Hollinshead M, et al. The organization of the human cerebral cortex estimated by intrinsic functional connectivity. J Neurophysiol. 2011;106:1125–65.

Article   PubMed Central   Google Scholar  

Rack-Gomer AL, Liu TT. Caffeine increases the temporal variability of resting-state BOLD connectivity in the motor cortex. NeuroImage. 2012;59:2994–3002.

McLachlan RS. A brief review of the anatomy and physiology of the limbic system. Can J Neurological Sci. 2009;36:S84–87.

Gramling L, Kapoulea E, Murphy C. Taste perception and caffeine consumption: an fMRI study. Nutrients. 2018;11:34.

Svenningsson P. Distribution, biochemistry and function of striatal adenosine A2A receptors. Prog Neurobiol. 1999;59:355–96.

Fredholm BB, Arslan G, Halldner L, Kull B, Schulte G, Wasserman W. Structure and function of adenosine receptors and their genes. Naunyn-Schmiedeberg’s Arch Pharmacol. 2000;362:364–74.

Adenosine FBB. Adenosine receptors and the actions of caffeine. Pharmacol Toxicol. 1995;76:93–101.

Mishina M, Ishiwata K. Adenosine receptor PET imaging in human brain. Int Rev Neurobiol. 2014;119:51–69.

Kahathuduwa CN, Dhanasekara CS, Chin SH, Davis T, Weerasinghe VS, Dassanayake TL, et al. l-Theanine and caffeine improve target-specific attention to visual stimuli by decreasing mind wandering: a human functional magnetic resonance imaging study. Nutr Res. 2018;49:67–78.

Brunyé TT, Mahoney CR, Lieberman HR, Taylor HA. Caffeine modulates attention network function. Brain Cogn. 2010;72:181–8.

Halassa MM, Kastner S. Thalamic functions in distributed cognitive control. Nat Neurosci. 2017;20:1669–79.

Baumann O, Borra RJ, Bower JM, Cullen KE, Habas C, Ivry RB, et al. Consensus Paper: the role of the cerebellum in perceptual processes. Cerebellum. 2015;14:197–220.

Richards G, Smith A. Caffeine consumption and self-assessed stress, anxiety, and depression in secondary school children. J Psychopharmacol. 2015;29:1236–47.

Kolesar TA, Bilevicius E, Wilson AD, Kornelsen J. Systematic review and meta-analyses of neural structural and functional differences in generalized anxiety disorder and healthy controls using magnetic resonance imaging. NeuroImage Clin. 2019;24:102016.

Koch SBJ, van Zuiden M, Nawijn L, Frijling JL, Veltman DJ, Olff M. Aberrant resting-state brain activity in posttraumatic stress disorder: a meta-analysis and systematic review: theoretical review: brain activity in PTSD during rest. Depress Anxiety. 2016;33:592–605.

Xu J, Van Dam NT, Feng C, Luo Y, Ai H, Gu R, et al. Anxious brain networks: a coordinate-based activation likelihood estimation meta-analysis of resting-state functional connectivity studies in anxiety. Neurosci Biobehav Rev. 2019;96:21–30.

Yacoubi ME, Ledent C, Parmentier M, Costentin J, Vaugeois JM. The anxiogenic-like effect of caffeine in two experimental procedures measuring anxiety in the mouse is not shared by selective A2A adenosine receptor antagonists. Psychopharmacology. 2000;148:153–63.

Yin Y-Q, Zhang C, Wang J-X, Hou J, Yang X, Qin J. Chronic caffeine treatment enhances the resilience to social defeat stress in mice. Food Funct. 2015;6:479–91.

Download references

This study was funded by the Institute for the Scientific Information on Coffee (ISIC) (ISIC_2017_NS); ISIC did not influence the experimental design or data analysis/interpretation. The laboratory was also supported by the project NORTE‐01‐0145‐FEDER000013 through the Northern Portugal Regional Operational Programme (NORTE 2020), under the Portugal 2020 Partnership Agreement, through the European Regional Development Fund (FEDER). RM, MP-P, and ME were supported by post-doctoral grants from the project ISIC_2017_NS. PSM was supported by a fellowship grant from the Fundação para a Ciência e a Tecnologia (FCT; grant number PDE/BDE/113601/2015) from the PhD-iHES program. RV was supported by a research fellowship of the project funded by FCT (UMINHO/BI/340/2018). AC was supported by a scholarship from the project NORTE-08-5639-FSE-000041 (NORTE 2020; UMINHO/BD/51/2017).

Author information

These authors contributed equally: Ricardo Magalhães, Maria Picó-Pérez, Madalena Esteves

Authors and Affiliations

Life and Health Sciences Research Institute (ICVS), School of Medicine, University of Minho, Campus de Gualtar, Braga, Portugal

Ricardo Magalhães, Maria Picó-Pérez, Madalena Esteves, Rita Vieira, Teresa C. Castanho, Liliana Amorim, Mafalda Sousa, Ana Coelho, Joana Cabral, Pedro S. Moreira & Nuno Sousa

ICVS/3B’s, PT Government Associate Laboratory, Braga/Guimarães, Portugal

Clinical Academic Center - Braga, Braga, Portugal

NeuroSpin, CEA, CNRS, Paris-Saclay University, Gif-sur-Yvette, France

Ricardo Magalhães

Center for Music in the Brain, Department of Clinical Medicine, Aarhus University, Aarhus, Denmark

Henrique M. Fernandes & Joana Cabral

Psychological Neuroscience Lab, CIPsi, School of Psychology, University of Minho, Braga, Portugal

Pedro S. Moreira

P5 Medical Center, Braga, Portugal

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Nuno Sousa .

Ethics declarations

Conflict of interest.

The authors declare no competing interests.

Ethical approval

The present study was conducted in accordance with the principles expressed in the Declaration of Helsinki (59th amendment) and was approved by the ethics committee of Hospital de Braga. All participants gave informed written consent after the study aims were explained and were informed of the option to withdraw from the study at any time.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplemental materials, rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Magalhães, R., Picó-Pérez, M., Esteves, M. et al. Habitual coffee drinkers display a distinct pattern of brain functional connectivity. Mol Psychiatry 26 , 6589–6598 (2021). https://doi.org/10.1038/s41380-021-01075-4

Download citation

Received : 19 December 2020

Accepted : 19 March 2021

Published : 20 April 2021

Issue Date : November 2021

DOI : https://doi.org/10.1038/s41380-021-01075-4

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

This article is cited by

Genome-wide association studies of coffee intake in uk/us participants of european ancestry uncover cohort-specific genetic associations.

• Hayley H. A. Thorpe
• Pierre Fontanillas
• Sandra Sanchez-Roige

Neuropsychopharmacology (2024)

Current coffee consumption is associated with decreased striatal dopamine transporter availability in Parkinson’s disease patients and healthy controls

• Minming Zhang

BMC Medicine (2023)

Perceived stress modulates the activity between the amygdala and the cortex

• Inês Caetano
• Sónia Ferreira

Molecular Psychiatry (2022)

• Sung Hoon Kang
• Jung Bin Kim

Scientific Reports (2021)

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings
• My Bibliography
• Collections
• Citation manager

Save citation to file

Email citation, add to collections.

• Create a new collection
• Add to an existing collection

Add to My Bibliography

Your saved search, create a file for external citation management software, your rss feed.

• Search in PubMed
• Search in NLM Catalog
• Add to Search

Coffee and health: a review of recent human research

Affiliation.

• 1 Linus Pauling Institute, Oregon State University, Corvallis, OR 97331, USA. [email protected]
• PMID: 16507475
• DOI: 10.1080/10408390500400009

Coffee is a complex mixture of chemicals that provides significant amounts of chlorogenic acid and caffeine. Unfiltered coffee is a significant source of cafestol and kahweol, which are diterpenes that have been implicated in the cholesterol-raising effects of coffee. The results of epidemiological research suggest that coffee consumption may help prevent several chronic diseases, including type 2 diabetes mellitus, Parkinson's disease and liver disease (cirrhosis and hepatocellular carcinoma). Most prospective cohort studies have not found coffee consumption to be associated with significantly increased cardiovascular disease risk. However, coffee consumption is associated with increases in several cardiovascular disease risk factors, including blood pressure and plasma homocysteine. At present, there is little evidence that coffee consumption increases the risk of cancer. For adults consuming moderate amounts of coffee (3-4 cups/d providing 300-400 mg/d of caffeine), there is little evidence of health risks and some evidence of health benefits. However, some groups, including people with hypertension, children, adolescents, and the elderly, may be more vulnerable to the adverse effects of caffeine. In addition, currently available evidence suggests that it may be prudent for pregnant women to limit coffee consumption to 3 cups/d providing no more than 300 mg/d of caffeine to exclude any increased probability of spontaneous abortion or impaired fetal growth.

PubMed Disclaimer

Similar articles

• Coffee: biochemistry and potential impact on health. Ludwig IA, Clifford MN, Lean ME, Ashihara H, Crozier A. Ludwig IA, et al. Food Funct. 2014 Aug;5(8):1695-717. doi: 10.1039/c4fo00042k. Food Funct. 2014. PMID: 24671262 Review.
• Coffee consumption and disease correlations. Gökcen BB, Şanlier N. Gökcen BB, et al. Crit Rev Food Sci Nutr. 2019;59(2):336-348. doi: 10.1080/10408398.2017.1369391. Epub 2017 Sep 29. Crit Rev Food Sci Nutr. 2019. PMID: 28853910 Review.
• Coffee consumption and human health--beneficial or detrimental?--Mechanisms for effects of coffee consumption on different risk factors for cardiovascular disease and type 2 diabetes mellitus. Ranheim T, Halvorsen B. Ranheim T, et al. Mol Nutr Food Res. 2005 Mar;49(3):274-84. doi: 10.1002/mnfr.200400109. Mol Nutr Food Res. 2005. PMID: 15704241 Review.
• The Impact of Coffee on Health. Nieber K. Nieber K. Planta Med. 2017 Nov;83(16):1256-1263. doi: 10.1055/s-0043-115007. Epub 2017 Jul 4. Planta Med. 2017. PMID: 28675917 Review.
• Coffee and liver diseases. Muriel P, Arauz J. Muriel P, et al. Fitoterapia. 2010 Jul;81(5):297-305. doi: 10.1016/j.fitote.2009.10.003. Epub 2009 Oct 13. Fitoterapia. 2010. PMID: 19825397 Review.
• Evaluation of the Caffeine Content in Servings of Popular Coffees in Terms of Its Safe Intake-Can We Drink 3-5 Cups of Coffee per Day, as Experts Advise? Wierzejska RE, Gielecińska I. Wierzejska RE, et al. Nutrients. 2024 Jul 23;16(15):2385. doi: 10.3390/nu16152385. Nutrients. 2024. PMID: 39125266 Free PMC article.
• Coffee Consumption and CYP1A2 Polymorphism Involvement in Type 2 Diabetes in a Romanian Population. Popa LC, Farcas SS, Andreescu NI. Popa LC, et al. J Pers Med. 2024 Jul 3;14(7):717. doi: 10.3390/jpm14070717. J Pers Med. 2024. PMID: 39063971 Free PMC article.
• Characterization of a new Camellia plant resource with low caffeine and high theobromine for production of a novel natural low-caffeine tea. Chen T, Ge Z, Yang X, Wang X, Zuo H, Liao Y, Chen Z, Zhang Z, Chen M, Zhao J, Luo J. Chen T, et al. Food Chem X. 2024 Jun 21;23:101586. doi: 10.1016/j.fochx.2024.101586. eCollection 2024 Oct 30. Food Chem X. 2024. PMID: 39036481 Free PMC article.
• Polyphenols vs. Caffeine in Coffee from Franchise Coffee Shops: Which Serving of Coffee Provides the Optimal Amount of This Compounds to the Body. Wierzejska RE, Gielecińska I, Hallmann E, Wojda B. Wierzejska RE, et al. Molecules. 2024 May 9;29(10):2231. doi: 10.3390/molecules29102231. Molecules. 2024. PMID: 38792093 Free PMC article.
• Coffee consumption might be associated with lower potential risk and severity of metabolic syndrome: national health and nutrition examination survey 2003-2018. Zhao H, Wang S, Han Y, Yao M, Zhang Y, Zeng X. Zhao H, et al. Eur J Nutr. 2024 Aug;63(5):1705-1718. doi: 10.1007/s00394-024-03367-1. Epub 2024 May 4. Eur J Nutr. 2024. PMID: 38703226

Publication types

• Search in MeSH

Related information

• Cited in Books
• PubChem Compound
• PubChem Compound (MeSH Keyword)
• PubChem Substance

LinkOut - more resources

Full text sources.

• Taylor & Francis

Other Literature Sources

• The Lens - Patent Citations

• Citation Manager

NCBI Literature Resources

MeSH PMC Bookshelf Disclaimer

The PubMed wordmark and PubMed logo are registered trademarks of the U.S. Department of Health and Human Services (HHS). Unauthorized use of these marks is strictly prohibited.

• Varieties Catalog
• Become a Member

Ensure the supply quality climate resilience future of coffee

Agricultural innovation.

We drive research to preserve origin diversity and secure future supply.

Farmers to Roasters

Our inclusive approach ensures innovations deliver shared value for farmers, origins, and industry.

A Sustainable Sector

Our members are committed to the long term sustainability of the coffee, its people and communities.

Our challenges

Innovation solves critical challenges and creates new and unimagined opportunities.

60% of global coffee supply comes from smallholder farmers. 1

Farmers who have access to tools, technology, and knowledge are more likely to be profitable and to stick with farming.

5 countries produce 74% of the world’s coffee 2

The world’s coffee supply is consolidating. Thriving origin diversity brings the economic benefits of coffee to farmers in more countries and gives roasters choices.

Coffee faces a $452 million/year innovation crisis 3

To prevent supply instability, consolidation, and loss we need $452 million more per year to be invested in coffee agricultural R&D.

We are facing supply shortages of millions of bags 4 as soon as 2030 5

Estimated demand for coffee is steadily increasing while climate change puts more pressure on landscapes. More productive varieties can help us meet the challenge.

We create the foundation for a vibrant and sustainable coffee sector.

Tools and resources

We develop global tools and resources to support farmers, technicians, breeders, roasters, and policymakers to drive impact at scale for millions of farmers.

Sensory Lexicon

Coffee varieties catalog, strategy 2021–2025, latest news, world coffee research’s f1 hybrids transition to pre-commercial trials, the future of robusta breeding: an update, world coffee research releases 2023 annual report, evaluating cup quality in global arabica variety trial reveals complex interaction between genetics and environment, wcr welcomes josué arevalo alvarenga, research technician for latin america, enhancing access to quality planting material, economic opportunities for coffee producers in uganda, world coffee research ceo jennifer “vern” long named 2024 notable leader in sustainability, first seedlings from the innovea network transplanted for field evaluation, world coffee research releases new report on seed quality assurance, underscores opportunities to strengthen sector, wcr welcomes dr. verônica belchior, research scientist for coffee quality evaluation, where to find us at the specialty coffee association expo 2024 in chicago, new, more complete arabica genome to allow for effective variety development.

Siles, P., Cerdán, C., & Staver, C. (2022). Smallholder Coffee in the Global Economy—A framework to explore transformation alternatives of traditional agroforestry for greater economic, ecological, and livelihood viability. Frontiers in Sustainable Food Systems, 6. https://doi.org/10.3389/fsufs.2022.808207.

The world's top five producers of coffee are Brazil, Vietnam, Colombia, Indonesia, and Ethiopia. This percentage was calculated using the total coffee production volume generated by the top five producing countries over the total global production volume for both arabica and robusta between 2019–2022. Source: International Coffee Organization trade statistics tables, https://www.ico.org/trade_statistics.asp.

Maredia, M. and Martínez, J. M. (2023). Coffee’s innovation crisis: Determining the size of the agricultural R&D investment gap for coffee amid growing consumer demand and the climate crisis. World Coffee Research. https://worldcoffeeresearch.org/resources/coffees-innovation-crisis

"Bags" refers to 60-kg bags.

Range calculated using historical coffee production data (1980-2020) and production projections under future demand scenarios and climate change scenarios. Source: Maredia, M. and Martínez, J. M. (2023). Coffee’s innovation crisis: Determining the size of the agricultural R&D investment gap for coffee amid growing consumer demand and the climate crisis. World Coffee Research. https://worldcoffeeresearch.org/resources/coffees-innovation-crisis

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• Sage Choice

Caffeine consumption and self-assessed stress, anxiety, and depression in secondary school children

Previous research suggests that effects of caffeine on behaviour are positive unless one is investigating sensitive groups or ingestion of large amounts. Children are a potentially sensitive subgroup, and especially so considering the high levels of caffeine currently found in energy drinks. The present study used data from the Cornish Academies Project to investigate associations between caffeine (both its total consumption, and that derived separately from energy drinks, cola, tea, and coffee) and single-item measures of stress, anxiety, and depression, in a large cohort of secondary school children from the South West of England. After adjusting for additional dietary, demographic, and lifestyle covariates, positive associations between total weekly caffeine intake and anxiety and depression remained significant, and the effects differed between males and females. Initially, effects were also observed in relation to caffeine consumed specifically from coffee. However, coffee was found to be the major contributor to high overall caffeine intake, providing explanation as to why effects relating to this source were also apparent. Findings from the current study increase our knowledge regarding associations between caffeine intake and stress, anxiety, and depression in secondary school children, though the cross-sectional nature of the research made it impossible to infer causality.

Introduction

Dose-dependent effects of caffeine on behaviour.

Short-term effects of caffeine consumption include enhanced mood and alertness ( Ferré, 2008 ; Kaplan et al., 1997 ; Lorist and Tops, 2003 ), improved exercise performance ( Doherty and Smith, 2004 ), increased blood pressure ( Riksen et al., 2009 ), improved ability to remain awake and mentally alert after fatigue ( Smit and Rogers, 2002 ), faster information processing speed and reaction time, and heightened awareness and attention ( Cysneiros et al., 2007 ). When consumed in moderation it appears that there are no serious adverse health effects associated with its use by adults ( Nawrot et al., 2003 ) or children ( Higdon and Frei, 2006 ; Mandel, 2002 ). However, it has been advised that those who are highly sensitive should not consume >400 mg/d, in order to avoid headaches, drowsiness, anxiety, and nausea ( Nawrot et al., 2003 ). A sensitive individual might experience adverse effects at a lower dose than less sensitive individuals. Children are often considered as sensitive individuals because of their size and developing central nervous system. This is concerning because many children and adolescents are frequent caffeine consumers (for instance, a recent US study found 73% of children to consume caffeine on a given day; Branum et al., 2014 ). It is important, therefore, to identify thresholds above which negative effects might occur. In the context of the current study, the thresholds in question relate to the group as a whole, with potential sensitivity to caffeine being defined by the participants being children.

The relatively recent introduction of ‘energy drinks’ to the consumer market has been highlighted as a cause for concern (e.g. Reissig et al., 2009 ). Energy drinks are soft drinks that manufacturers claim boost performance and endurance ( Meadows-Oliver and Ryan-Krause, 2007 ), with the main active ingredient being caffeine ( McLellan and Lieberman, 2012 ). These products are often strategically marketed towards the young consumer ( Reissig et al., 2009 ), with 30–50% of adolescents and young adults now known to consume them ( Seifert et al., 2011 ). Energy drinks have also been associated with behavioural problems ( Richards et al., 2015a ), and a number of serious health complications ( Reissig et al., 2009 ).

A potential avenue by which energy drink use may negatively affect health is through their association with risk-taking behaviours (see Arria et al., 2014 ). Miller (2008a) , for instance, reported that the frequency of energy drink consumption in US undergraduates was positively associated with smoking, drinking, alcohol problems, use of illicit prescription drugs and marijuana, sexual risk-taking, fighting, seatbelt omission, and taking risks on a dare. However, it should be noted that such effects might also be explainable by personality characteristics of high users of energy drinks (for example, adherence to a ‘toxic jock’ identity; Miller, 2008b ), rather than necessarily to the products themselves.

Another potential route that energy drinks may negatively affect health is through caffeine’s capacity to disrupt sleep. Energy drink use has been associated with daytime sleepiness and weekly ‘jolt and crash’ episodes ( Kristjánsson et al., 2011 ; Malinauskas et al., 2007 ), though the products also appear to be used to counter the effects of insufficient sleep ( Malinauskas et al., 2007 ). Although findings such as these may implicate energy drinks in particular, Kristjánsson et al. (2013) have reported that caffeine consumption itself is positively associated with self-reported violent behaviour and conduct disorder. Furthermore, James et al. (2011) observed a strong inverse relationship between caffeine intake and academic attainment, 32% of which was explained by mediating effects of daytime sleepiness and other licit substance use. Due to findings such as these it is considered to be of particular importance to investigate the effects of caffeine from difference sources, as well as its overall intake.

Associations between caffeine intake and stress, anxiety, and depression

The consumption of caffeinated beverages is known to be a coping strategy used by college students in the management of stressful academic situations ( Lazarus, 1993 ; Thoits, 1995 ), with 49% of a representative stratified sample of Puerto Rican students reporting caffeinated products to be useful for coping with stress ( Ríos et al., 2013 ). Pettit and DeBarr (2011) have also reported a positive relationship between energy drink consumption and perceived stress levels in undergraduate students. Though the use of caffeine is moderately related to a range of psychiatric and substance use disorders in the general population, the relationships appear not to be causal ( Kendler et al., 2006 ), and results between studies are equivocal (for a review of the area see Lara, 2010 ). Discerning the nature and direction of relationships between such variables becomes even more difficult when considering the self-medication hypothesis (e.g. Khantzian, 1997 ). The idea here is that people may self-medicate with legal and/or illicit substances, with evidence having already been provided to suggest that some individuals with mental health problems use caffeinated energy drinks for such purposes ( Chelben et al., 2008 ).

In some cases positive effects of caffeine have been observed. For instance, low doses have been shown to reduce anxiety and elevate mood ( Haskell et al., 2005 ; Lieberman et al., 1987 , 2002 ; Smith, 2009a ; Smith et al., 1999 ). Smith (2009b) also reported that caffeine consumption was associated with reduced risk of depression compared with non-consumption in a population study.

Negative effects of caffeine on stress and mental health have also been observed. Gilliland and Andress (1981) , for instance, reported higher anxiety levels in moderate and high caffeine consumers compared with abstainers in a student sample. Case reports also suggest that mania can be induced by a high intake of caffeine ( Ogawa and Ueki, 2003 ) or energy drinks ( Sharma, 2010 ). These results are supported by the finding of Kaplan et al. (1997) , that 250 mg of caffeine can increase elation in healthy volunteers, whereas 500 mg increases irritability. Other studies, however, have reported null findings. James et al. (1989) , for instance, found no relationships between caffeine intake and anxiety or depression in medical students.

In the general population, negative effects of caffeine are usually observed in relation to excessive intake. At extremely high doses its consumption can induce a condition known as ‘caffeinism’. Symptoms include anxiety, nervousness, restlessness, insomnia, excitement, psychomotor agitation, dysphoria, and a rambling flow of thoughts and speech ( Gilliland and Andress, 1981 ; Greden, 1974 ), which have been considered to mimic a clinical picture known as ‘mixed mood state’ ( Lara, 2010 ).

Larger effects of caffeine seem to occur in sensitive individuals, with psychiatric patients appearing to make up one such group. Higher sensitivity to the anxiogenic effects of high doses (typically >400 mg), for instance, has been observed in patients with panic disorder ( Boulenger et al., 1984 ; Charney et al., 1985 ), generalised panic disorder ( Bruce et al., 1992 ), and to a lesser extent, depression ( Lee et al., 1988 ). Similar findings have also been made in patients with performance social anxiety disorder (though not generalised social anxiety disorder; Nardi et al., 2009 ), and excessive intake may interfere with the recovery of patients with bipolar disorder and manic-type mood episodes ( Caykoylu et al., 2008 ; Dratcu et al., 2007 ; Tondo and Rudas, 1991 ).

Another potentially sensitive subgroup is that of young consumers. Certain psychiatric symptoms appear to occur at an alarming rate in this group. For example, the prevalence of major depressive disorder is known to range from 0.4% to 8% in adolescents ( Birmaher et al., 1996 ; Fleming and Offord, 1990 ; Roberts et al., 1995 ), with approximately 30% reporting at least one current symptom of a major depressive episode ( Roberts et al., 1995 ). Depressive symptoms have also been found to correlate positively with coffee consumption in middle- and high-school students ( Fulkerson et al., 2004 ), and positive associations with the Children’s Depression Inventory have been reported in both children and adolescents ( Luebbe and Bell, 2009 ). However, as with research in adults, some studies have also reported null findings. Luebbe and Bell (2009) , for instance, found no relationship between anxiety and caffeine in children and adolescents.

Aims of the current research

The general lack of research relating to the effects of caffeine on stress, anxiety, and depression in children is an area that the current paper will try to address. In order to do this, the Diet and Behaviour Scale (DABS; Richards et al., 2015b ), a measure of intake of food and drinks (including caffeinated products) that may affect psychological outcomes, was administered to a large cohort of secondary school children from the South West of England. The current paper used the DABS for two purposes: (1) to provide estimates of weekly caffeine intake from energy drinks, cola, tea, and coffee, and (2) so that additional aspects of diet could be controlled for in multivariate analyses.

Along with the DABS, single-item measures of self-assessed stress, anxiety, and depression were administered. Single items were chosen as they have been shown to be valid and reliable, allowing for the identification of overall risk whilst reducing the time costs associated with administering multi-item measures ( Williams and Smith, 2012 ). The items themselves came from the Wellbeing Process Questionnaire ( Williams, 2014 ), have been validated against full measures, demonstrated to correlate well, and appear to be as sensitive as the full-length measures with which they were compared ( Williams, 2015 ; Williams and Smith, 2013 ).

It was hypothesised that high consumption of caffeine would be associated with high stress, anxiety, and depression, and that such relationships would not be dependent on the source from which caffeine was obtained. However, as no interventions were conducted, and data presented here are only cross-sectional in nature, it should be acknowledged that it is not possible to infer causality or the direction of relationships observed.

The Cornish Academies Project was a large-scale longitudinal programme of research designed to investigate dietary effects on school performance, general health, and stress, anxiety, and depression in secondary school children. Two cross-sections of data were collected from three academies in the South West of England. The first cross-section (T1) was collected 6 months prior to the second (T2). The current paper presents analyses using data from the latter cross-section only, as information relating to stress, anxiety, and depression were not collected at the former.

Participants

In total, 3071 secondary school pupils were asked to take part in the Cornish Academies Project at T1; 2610 (85%) agreed to participate. At T2, the cohort consisted of 3323 pupils, and 2307 completed the questionnaires. A relatively balanced sex ratio (48.5% male, 51.5% female), and an age range of 11–17 ( M = 13.6, SD = 1.49) were observed (for a more detailed description of the sample see Richards et al., 2015b ).

Apparatus/materials

The DABS ( Richards et al., 2015b ) is a 29-item questionnaire developed for the purpose of assessing intake of common dietary variables with an onus on functional foods, and foods and drinks of current concern. The DABS contains 18 questions that assess frequency of consumption on a five-point scale (1 = never, 2 = once a month, 3 = once or twice a week, 4 = most days [3–6], 5 = every day), and 11 questions to assess amounts typically consumed. It has been associated with a four-factor structure in secondary school children labelled Junk Food, Caffeinated Soft Drinks/Gum, Healthy Foods, and Hot Caffeinated Beverages (see Richards et al., 2015b ).

Because caffeine content is known to vary considerably between energy drink products ( Reissig et al., 2009 ), participants were asked to state the brand names of those that they consumed. This measure was included in order to increase the accuracy of estimating caffeine consumption. In addition to this, as diet may reflect general lifestyle (e.g. Akbaraly, 2009 ), five further questions were administered. Three items were used to gauge exercise frequency (mildly energetic, moderately energetic, and vigorous), with answers being given on a four-point scale (1 = three times a week or more, 2 = once or twice a week, 3 = about once to three times a month, 4 = never/hardly ever). In addition to this, participants were asked to state how many hours per night they typically spent asleep, and to give an indication of how good they perceived their general health to have been over the previous 6 months (1 = very good, 2 = good, 3 = fair, 4 = bad, 5 = very bad). Participants were then asked to state how frequently they had experienced stress, anxiety, and depression over the previous 6 months, on a five-point scale (1 = not at all, 2 = rarely, 3 = sometimes, 4 = frequently, 5 = very frequently), though no clinical evaluations were made. No further descriptions of ‘stress’, ‘anxiety’, or ‘depression’ were provided as it was assumed that participants would understand the concepts at hand.

Design and procedure

Schoolteachers administered the DABS as well as the lifestyle, stress, anxiety, and depression questions to the pupils at their respective academies. Demographic information was acquired through the School Information Management System (SIMS) and stored in a confidential database in Cardiff. This information included age, sex, school attendance, number of detentions/behavioural points received, English and maths attainment at Key Stage 3/Key Stage 4, school year, ethnicity, presence/absence of a special educational needs (SEN) status, eligibility/ineligibility to receive free school meals (FSM; a proxy indication of socioeconomic status; Shuttleworth, 1995 ), whether or not English was spoken as an additional language, and whether or not children were cared for by a non-parental guardian.

All questionnaire and demographic data were anonymised prior to being merged into a single database. Ethical clearance was granted by Cardiff University’s School of Psychology Ethics Committee, and informed consent was acquired from all participants (as well as their parents) before data were collected. All data analysis was conducted using IBM SPSS Statistics Version 20.

Statistical analysis

The representativeness of the sample was investigated by comparing SIMS data for those who completed the questionnaires with that of those who did not, though frequency data relating to stress, anxiety, and depression are not provided here because they have already been reported elsewhere (see Richards and Smith, 2015 ). Weekly caffeine consumption was calculated from the DABS items used to measure the amount of consumption of energy drinks, cola, tea, and coffee. Linear-by-linear trends were then investigated between total weekly caffeine intake and stress, anxiety, and depression, and were followed up with binary logistic regression analyses (using the ‘enter’ method), so that additional variance from diet, demography, and lifestyle could be controlled for statistically. In order to investigate interactions between caffeine use and sex, multivariate analyses were conducted for males and females separately. It was also deemed important to examine the effects of each individual source of caffeine (i.e. that consumed specifically from energy drinks, cola, tea, and coffee). As with the analyses of total weekly caffeine intake, these effects were initially investigated using linear-by-linear trends, and then with binary logistic regression to control for additional covariates (though in this instance separate analyses were not conducted for males and females).

Demographic and lifestyle variables

Considerable variance in demographic background and lifestyle was observed within the sample; for frequency data, see Table 1 . Participants’ average number of sleep hours and frequency of exercise (a single-factor analysed variable derived from items measuring mild, moderate, and vigorous exercise) that are used as control variables in the current study have been described elsewhere (see Richards et al., 2015b ).

Frequency information for demographic variables.

Representativeness of the sample

A relatively high response rate of 88.4% was observed for completion of the DABS. To investigate how representative the sample was in reference to the academies from which it came, Chi-square tests were conducted to determine if SIMS data for those who completed the DABS differed from SIMS data of those that did not. It should be noted that a similar analysis presented in Richards et al. (2015b) relates to T1 from the Cornish Academies Project, whereas that presented here relates to T2.

It was found that the academy a pupil came from was significantly related to their likelihood of responding to the questionnaires, with Academy 1 and Academy 3 providing fewer respondents, and Academy 2 providing more respondents, than expected, χ2 (2, N = 3323) = 241.172, p < .001. The school year that a participant came from was also related to their likelihood of completing the questionnaires, χ2 (4, N = 3049) = 34.681, p < .001. A significant linear-by-linear association was observed, with the likelihood of responding being negatively associated with school year, χ2 (1, N = 3049) = 30.245, p < .001. Children with a SEN status were also less likely to answer the questionnaire, χ2 (1, N = 3083) = 23.142, p < .001, as were children who were eligible to receive FSM, χ2 (1, N = 3049) = 25.116, p < .001.

Associations between total weekly caffeine intake and stress, anxiety, and depression

Univariate associations between total weekly caffeine intake and stress, anxiety, and depression.

Single items from the DABS were used to estimate weekly caffeine intake, with the following values being assigned: cup of coffee (80 mg), cup of tea (40 mg), can of cola (25 mg), can of energy drink (133 mg). The values used for coffee, tea, and cola, were based on updated versions of those reported by Brice and Smith (2002) , which were themselves based on values provided by Barone and Roberts (1996) and Scott et al. (1989) ; the value used for energy drinks was the mean caffeine content of the three brands most commonly reported by the current sample (which together accounted for 53.2% of all cases). Caffeine totals consumed from coffee, tea, energy drinks, and cola were then added together to create a variable for total weekly consumption. It was found that caffeine intake was higher in males than females, both in total amount, as well as in that consumed from energy drinks, cola, and coffee (though there was no difference regarding caffeine consumed from tea; for descriptive statistics, see Table 2 ). Total weekly caffeine was subsequently recoded into a categorical variable consisting of the following six consumption groups: 0 mg/w, 0.1–250 mg/w, 250.1–500 mg/w, 500.1–750 mg/w, 750.1–1000 mg/w, >1000 mg/w.

Descriptive statistics and sex differences for self-reported stress, anxiety and depression, and weekly caffeine intake as calculated from the DABS.

Self-assessed stress, anxiety, and depression were all found to be significantly higher in females compared with males (for descriptive statistics, see Table 2 ). The single-item measures were then dichotomised, with those answering with 1 or 2 (‘not at all’ or ‘rarely’ experienced stress, anxiety, or depression) making up the above average mental health group, and those answering with 3, 4, or 5 (‘sometimes’, ‘frequently’, or ‘very frequently’ experienced stress, anxiety, or depression) making up the below average mental health group.

Linear-by-linear associations were investigated between the dichotomous variables for stress, anxiety, and depression, and the categorical variable created from total weekly caffeine intake. The analysis found the >1000 mg/w condition to be associated with high stress, anxiety, and depression. In addition to this, consuming 0.1–250 mg/w was associated with low stress, and non-consumption was associated with low depression, though the latter effect was not significant. For linear-by-linear associations and cross-tabulations between total weekly caffeine intake and stress, anxiety, and depression, see Table 3 .

Cross-tabulations between total weekly caffeine intake and stress, anxiety, and depression.

Note. Mean weekly caffeine intake for each consumption group was as follows: 0 mg M = 0 mg ( SD = 0), 0.1–250 mg/w M = 117.83 ( SD = 69.32), 250.1–500 mg/w M = 355.94 ( SD = 70.61), 500.1–750 mg/w M = 616.37 ( SD = 69.99), 750.1–1000 mg/w M = 865.09 ( SD = 72.71), >1000 mg/w M = 1651.74 ( SD = 750.33).

Multivariate associations between total weekly caffeine intake and stress, anxiety, and depression

The analyses described in the previous section indicate that being a very high consumer of caffeine is a predictor of high levels of stress, anxiety, and depression. It was therefore deemed important to further investigate such effects at the multivariate level, so that additional variance could be controlled for statistically. In order to do this, binary logistic regression analyses (using the ‘enter’ method) were conducted upon the dependent variables of stress, anxiety, and depression. The same categorical variable for total weekly caffeine intake described in the previous section was used, and the non-consumption group was set as the comparison. The additional covariates entered were diet (the DABS subscale scores for Junk Food and Healthy Foods; Caffeinated Soft Drinks/Gum and Hot Caffeinated Beverages were not entered as they were comprised of caffeinated products; for a description of these variables see Richards et al., 2015b ), demography (sex, school, school year, presence/absence of a SEN status, and the eligibility/ineligibility to receive FSM), and lifestyle (sleep hours, exercise frequency, and school attendance). It was, however, deemed inappropriate to attempt to control for ethnicity, whether English was spoken as an additional language, and whether or not the child was cared for by a non-parental guardian, due to the numbers present in the minority groups being particularly small.

After controlling for covariates, the overall effect of caffeine on stress was not significant, Wald = 6.252, p = .283, and none of the consumption groups differed from the non-consumption group. However, total weekly caffeine intake remained a significant predictor of anxiety, Wald = 12.39, p = .03. This effect reflected increased risk of high anxiety occurring in the >1000 mg/w group, though none of the other conditions differed significantly from the non-consumers. For odds ratios and 95% confidence intervals for the multivariate association between total weekly caffeine intake and anxiety, see Figure 1 .

Odds ratios and 95% confidence intervals for multivariate associations between total weekly caffeine intake and anxiety.

The effect of caffeine on depression also remained significant after controlling for covariates, Wald = 14.682, p = .012. In this case increased risk was associated with each of the consumption groups compared with the non-consumers (though the effect relating to the 250.1–500 mg/w group was only marginally significant, and the effect relating to the 500.1–750 mg/w group was not significant). For odds ratios and 95% confidence intervals for the multivariate associations between caffeine and depression, see Figure 2 .

Odds ratios and 95% confidence intervals for multivariate associations between total weekly caffeine intake and depression.

Sex differences in associations between total weekly caffeine intake and stress, anxiety, and depression

Due to the large sample size available, and because sex differences in responses to caffeine in adolescents have been reported (e.g. Temple and Ziegler, 2011 ), it was deemed meritorious to investigate interactions between sex and caffeine intake. To do this, the same methodology outlined in the previous section was used (i.e. binary logistic regression analyses were conducted, and the same covariates were entered), except that the caffeine*sex interaction term was included instead of the main effects of caffeine and sex. Significant interactions were observed for each of the outcome variables: stress, Wald = 31.927, p < .001, anxiety, Wald = 50.341, p < .001, depression, Wald = 45.038, p < .001.

In order to further investigate the interactions between sex and caffeine intake on stress, anxiety, and depression, separate multivariate analyses were conducted in males and females. The overall effect of caffeine on stress was not significant in males, Wald = 5.193, p = .393, or females, Wald = 4.243, p = .515, though males who consumed >1000 mg/w were marginally more likely to report high stress compared with controls, OR = 1.891, 95% CI [.943, 3.792], p = .073. The effect of caffeine on anxiety was not significant in females, Wald = 8.307, p = .14, and none of the consumption groups differed significantly from the control. In males, however, the effect was significant, Wald = 13.186, p = .022. This reflected increased risk of high anxiety in the 0.1–250 mg/w, 250.1–500 mg/w, and >1000 mg/w conditions, with the effect being most apparent in the last condition. For odds ratios and 95% confidence intervals for the multivariate association between total weekly caffeine intake and anxiety in males see Figure 3 .

Odds ratios and 95% confidence intervals for multivariate associations between total weekly caffeine intake and anxiety in males.

The overall effect of caffeine on depression in males was not significant, Wald = 7.882, p = .163. However, each caffeine consumption group was associated with increased risk compared with the control (though the effect relating to the 500.1–750 mg/w group was only marginally significant, and the effect in the 750.1–1000 mg/w condition was not significant). The overall effect in females was significant, Wald = 13.137, p = .022, and reflected increased risk in both the 750.1–1000 mg/w and >1000 mg/w groups. For odds ratios and 95% confidence intervals for the multivariate associations between total weekly caffeine intake and depression in males and females, see Figures 4 and ​ and5, 5 , respectively.

Odds ratios and 95% confidence intervals for multivariate associations between total weekly caffeine intake and depression in males.

Odds ratios and 95% confidence intervals for multivariate associations between total weekly caffeine intake and depression in females.

Associations between individual caffeine sources and stress, anxiety, and depression

Univariate associations between individual caffeine sources and stress, anxiety, and depression.

In order to determine whether the source of caffeine was important regarding the relationships reported in the previous section, caffeine from energy drinks, cola, tea, and coffee were recoded into three groups (non-consumption, low consumption, and high consumption), and linear-by-linear associations were investigated in relation to stress, anxiety, and depression. Because the distributions were skewed, the cut-off points to define what constituted ‘low consumption’ and ‘high consumption’ were determined in a manner that assigned relatively balanced numbers of participants to each group. These distinctions are shown in Table 4 ; essentially ‘low consumption’ related to one can of energy drink, one can of cola, two cups of coffee, and three cups of tea per week, and ‘high consumption’ related to any values in excess of these.

Cross-tabulations between weekly caffeine intake from energy drinks, cola, coffee, and tea, and stress, anxiety, and depression.

Note. Caffeine amounts listed relate to weekly consumption.

Caffeine consumed from energy drinks and tea was not associated with stress, anxiety, or depression. Interestingly, although consumption of caffeine from cola was not associated with anxiety or depression, its non-consumption was associated with high stress levels, and being a low consumer was associated with low stress levels.

Positive linear relationships were observed between caffeine consumption from coffee and stress, anxiety, and depression (for linear-by-linear associations and cross-tabulations between stress, anxiety, and depression, and caffeine consumed from individual sources, see Table 4 ). However these associations are likely explained by coffee being the major contributor to high overall caffeine intake. This is reflected in the observation that those above the median for caffeine intake from coffee consumed more total caffeine than did those above the median for each of the other sources: caffeine from coffee low M = 261.42 ( SD = 331.82), high M = 827.65 ( SD = 748.51); caffeine from energy drinks low M = 247.63 ( SD = 382.38), high M = 674.24 ( SD = 649.38); caffeine from tea low M = 225.97 ( SD = 365.43), high M = 640.55 ( SD = 633.11); caffeine from cola low M = 295.12 ( SD = 448.63), high M = 486.88 ( SD = 585).

Multivariate associations between individual caffeine sources and stress, anxiety, and depression

In order to further investigate associations between caffeine from different sources and stress, anxiety, and depression, the non-consumption/low consumption/high consumption variables for caffeine from energy drinks, cola, tea, and coffee were entered together into binary logistic regression analyses using the ‘enter’ method. The same dietary, demographic, and lifestyle variables that were controlled for in the multivariate analyses of total weekly caffeine intake were again entered as covariates here.

Low consumption of caffeine from energy drinks was associated with high stress, though the overall effect was not significant. Both low and high consumption of caffeine from cola, on the other hand, were significantly associated with low stress. Low caffeine from energy drinks and high caffeine from coffee were both marginally associated with high anxiety, though neither effect was significant overall. Low consumption of caffeine from tea was associated with high depression, and the overall effect was significant. High caffeine consumption from coffee was also associated with high depression, though in this case the overall effect was not significant. For odds rations, 95% confidence intervals, and p -values for all multivariate level associations between individual caffeine sources and stress, anxiety, and depression, see Table 5 .

Multivariate associations between individual sources of caffeine and stress, anxiety, and depression.

The current study aimed to present cross-sectional data from the Cornish Academies Project to investigate associations between caffeine consumption and stress, anxiety, and depression in secondary school children. Based on findings from the literature it was predicted that excessive caffeine intake would be associated with high stress, anxiety, and depression, and that such effects would not be dependent on any particular source of caffeine. In addition to this, separate analyses were conducted in males and females in order to investigate interactions between caffeine and sex.

Relationships between total weekly caffeine intake and stress, anxiety, and depression

Initial positive relationships were observed between total weekly caffeine intake and stress, anxiety, and depression. After adjusting for dietary, demographic, and lifestyle covariates, the effect on stress disappeared. However, consuming >1000 mg/w remained a predictor of high anxiety, and caffeine consumption in general appeared to be associated with higher instances of depression compared with non-consumption (although the effect was also most pronounced in those who consumed >1000 mg/w).

Though the above findings mainly replicated those reported in adults (e.g. Gilliland and Andress, 1981 ; Pettit and DeBarr, 2011 ), the effects appeared to occur at lower doses, which is most likely a reflection of the lower bodyweight of children compared with adults. One finding that differed considerably from those made in adult populations was that of depression. Smith (2009b) observed caffeine consumption to be beneficial compared with its abstinence, whereas the opposite pattern of results was observed here. This finding is therefore likely to highlight differences between the populations studied.

As significant interactions between total weekly caffeine intake and sex were observed in relation to each of the outcome variables, separate multivariate analyses were conducted for males and females. No association between caffeine and anxiety appeared in females; in males, higher instances of anxiety occurred in the 0.1–250 mg/w, 250.1–500 mg/w, and >1000 mg/w conditions, with the largest effect occurring in the last group. For depression, effects occurred in both males and females. In males, increased risk was associated with each group that consumed caffeine compared with non-consumers (though consuming 500.1–750 mg/w was only a marginally significant predictor, and consuming 750.1–1000 mg/w was not significantly related). In females, consuming either 750.1–1000 mg/w or >1000 mg/w was significantly associated with higher reporting of depression. These observations are consistent with other findings, such as caffeine having been shown to produce greater arousal effects in young males compared with females ( Adan et al., 2008 ), and to have a higher propensity for reinforcement in adolescent males compared with females ( Temple et al., 2009 ). Though it may be that male adolescents are more vulnerable to harmful effects of caffeine than are females, these results may also reflect sexually dimorphic personality characteristics, or the observation that overall caffeine consumption in the current study was higher in males than in females.

When individual caffeine sources were investigated, negative effects were observed in relation to coffee, tea, and energy drinks, though they were not consistent across variables and often only marginally statistically significant. One relationship of particular interest was however observed: both low (0.1–25 mg/w) and high (>25 mg/w) levels of caffeine consumed from cola were associated with low stress. This finding may reflect reports of students using caffeinated products to cope with stress (e.g. Ríos et al., 2013 ). However, the general lack of consistent findings from this analysis as a whole suggests that, when investigating its effects on stress, anxiety, and depression, caffeine is best examined in terms of total intake rather than by differentiating between individual sources.

Methodological limitations and directions for future research

Though the current study has addressed a gap in the literature, several methodological limitations should be considered when interpreting the findings. One such limitation is that the participants who completed the questionnaires were not fully representative of the schools from which they came. However, taking a multivariate approach to data analysis in which demographic and lifestyle variables could be controlled for statistically is deemed to have been an effective method for addressing this issue. Nevertheless, as the population studied came from a very specific demographic group (i.e. 11–17-year-old White children from the South West of England), further research is needed that focuses on more representative samples.

Another limitation of the current research was that the chronicity of caffeine use was not taken into account. For instance, a weekly cycle of caffeine use in adolescents was reported by Pollak and Bright (2003) , in which consumption peaked during the weekend (Saturday), and was lowest in the middle of the week (Wednesday). Coupled with the observations that adolescents sometimes use caffeinated products to delay sleep onset (e.g. Calamaro et al., 2009 ) and to counteract the effects of sleepiness during the day ( Malinauskas et al., 2007 ), it is possible that the timing of administration of the questionnaire may have been of importance.

A further limitation of the current study is that it utilised a cross-sectional design. This means that all effects observed here are correlational, and that causation cannot be inferred. Therefore the possibility of reverse-causation, or indeed bi-directionality, cannot be disregarded. For instance, high caffeine consumption may be a cause of high stress, anxiety, and depression, but suffering from such conditions may also lead towards the high consumption of caffeinated products as a coping strategy. Furthermore, it is possible that the effects observed here are attributable to personality characteristics associated with caffeine users, rather than to their use of caffeine. Future research should therefore aim to conduct intervention studies in order to investigate the nature of these relationships further.

Conclusions

The current study has presented results that suggest caffeine consumption may be associated with stress, anxiety, and depression in secondary school children, though the effect on stress disappeared after additional dietary, demographic, and lifestyle variance was controlled for statistically. The effects observed also appeared to differ between males and females. Though caffeine consumption was associated with anxiety in males at the multivariate level, no such observation was made in females. Furthermore, though the effects relating to depression occurred in both sexes, the threshold at which they appeared was lower in males than it was in females.

Initial analyses of individual caffeine sources implied that coffee may have been responsible for the effects observed in relation to total caffeine intake, but further investigations suggested this not to have been the case, and that they were likely attributable to caffeine consumption in general rather than to any particular source. The study also identified very high caffeine intake (>1000 mg/w) to be a risk factor associated with anxiety and depression, although effects were sometimes detected at lower doses. These findings may therefore be a concern for public health and school policy, and should be considered an important area for further investigation.

Acknowledgments

The authors would like to acknowledge the contribution of The Waterloo Foundation for funding the research. In addition, the authors wish to express their gratitude for the on-going support and collaboration with Pool Academy, Penrice Community College, and Treviglas Community College, as well as to thank each of the teachers and pupils who made the research possible.

Declaration of Conflicting Interests: The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding: The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The current research was supported by a grant from The Waterloo Foundation (grant number: 503692).