Worldwide, a high proportion of adults drink coffee daily, making it one of the most popular beverages globally. Coffee contains a range of bioactive compounds, including caffeine, chlorogenic acid, polyphenols and small amounts of vitamins and minerals (Spiller, 1984). Epidemiological studies suggest coffee has beneficial effects on various conditions including stroke (Larsson and Orsini, 2011), heart failure (Mostofsky et al., 2012), cancers (Wang et al., 2016), diabetes (Akash et al., 2014), and Parkinson’s disease (Ross et al., 2000). Alzheimer’s disease (AD) is a neurodegenerative disease characterized by progressive impairment of learning, memory and other cognitive deficits, with extracellular deposition of Aβ-amyloid (Aβ) protein within the brain leading to neuroinflammation, synaptic loss and neuronal death (Villemagne et al., 2013). Several studies suggest a protective role of coffee, with reduced risk of mild cognitive impairment (MCI) and AD reported (van Gelder et al., 2007; Eskelinen et al., 2009; Arab et al., 2011; Liu et al., 2016; Wierzejska, 2017; Wu et al., 2017). However, there are limited longitudinal data from cohorts of cognitively normal older adults describing associations of coffee consumption with distinct domains of cognition, and concurrently investigating potential neuropathological mechanisms underpinning any such associations.

Results from a meta-analysis conducted in 2016, which included nine published prospective cohort studies, found drinking one to two cups of coffee daily was associated with lower incidence of cognitive disorders (i.e., cognitive decline, cognitive impairment, AD, and all cause dementia) compared with less than one cup; studies ranged in follow-up from 1.3 to 28 years (Liu et al., 2016). Moreover, a recent cross-sectional analysis found self-reported lifetime intake of two or more cups of coffee per day was associated with lower rates of “Aβ positivity” (presence of significant brain Aβ), compared to less than two cups per day, in 411 non-demented older adults. However, current coffee intake was not related to “Aβ positivity” in this cohort, and neither current, nor lifetime intake, was related to cerebral mean cortical thickness or white matter hyperintensity volume (Kim et al., 2019). A second study investigating the effect of self-reported coffee consumption on brain volume found both high coffee consumption (≥ 4 cups per day) and low coffee consumption (no coffee or < 1 cup per day) to be associated with larger hippocampal volume cross-sectionally (p for test for quadratic trend = 0.001). There were no associations with volumes of the neocortex and striatum (accumbens, putamen, globus pallidus) or total intracranial volume. The cohort was homogeneous in nature, with all participants young Caucasian females (aged 23.2 ± 2.7 years), right handed with university education, and no history of smoking or drug/alcohol abuse (Perlaki et al., 2011). Conversely, higher self-reported coffee consumption (>3 cups/day compared to 0–1 cup/day) has also been associated cross-sectionally with smaller hippocampal volume, in 2914 individuals aged 59.28 ± 7.2 years (p = 0.033) (Araujo et al., 2016). However, this association was only observed when body mass index, previous coronary heart disease, alcohol consumption, and smoking were included in the statistical model, along with age, sex, and educational attainment (Araujo et al., 2016). Well-designed longitudinal cohort studies are necessary to further understand the influence of coffee consumption on AD phenotype and rates of decline in different cognitive domains.

The aim of the current study was to investigate the relationship of self-reported habitual coffee intake to rates of decline in varying cognitive domains, assessed using a comprehensive neuropsychological battery, over 126 months in 227 older adults classified as cognitively normal at baseline. Furthermore, we aimed to investigate whether habitual coffee intake was associated with rates of cerebral Aβ-amyloid deposition, or brain volume atrophy over the same time-period, in subsets of 60 and 51 participants, respectively. These investigations were conducted using data from the well-characterized Australian Imaging, Biomarkers and Lifestyle study of ageing (AIBL) (Fowler et al., 2021).

The current cohort comprised 227 participants (40.5% male) with an average age of 69.7 years at baseline. Almost 27% were carriers of the APOE ε4 allele (genotypes ε2/4, ε3/4, or ε4/4) and nearly 50% had 12 or less years of education. Of the 60 participants included in the PET imaging analysis, 46.7% were male with an average age of 70.44 years, and almost 40% were carriers of the APOE ε4 allele. Of the 51 participants included in the MRI analysis, all were also included in the PET imaging analysis (supplemented by an additional n = 9 who did not undergo MRI). Forty-nine percent were male with an average age of 69.92 years, and over 43% were APOE ε4 carriers (Table 1).

Habitual coffee intake was positively associated with the cognitive domains of executive function (p < 0.01), attention (p < 0.05), and the AIBL PACC (p < 0.01), such that higher coffee consumption was associated with slower cognitive decline in these domains over 126 months (Table 2). Within the executive function, attention and AIBL PACC models, the following covariates contributed to variance in the outcome: for executive function, APOE ε4 allele carriage (F = 4.421, p = 0.037); for attention, age, and education (F = 4.779, p = 0.030; F = 6.335, p = 0.013, respectively); for the AIBL PACC, sex, age, and education (F = 8.987, p = 0.003; F = 30.131, p < 0.001; F = 20.655, p < 0.001, respectively). Moreover, odds ratios showed that higher coffee consumption was associated with lower risk of converting from cognitively normal to MCI or AD status over 126 months (β = −0.004; p = 0.006).

Daily coffee intake was also associated with cerebral Aβ-amyloid burden, such that higher coffee consumption was associated with slower Aβ-amyloid accumulation over 126 months (p < 0.05; Table 2). Within the model, the following covariates contributed to variance in the outcome: energy intake, and time from baseline to FFQ completion (F = 4.860, p = 0.030; F = 4.865, p = 0.034, respectively). Furthermore, odds ratios showed that higher coffee consumption was associated with lower risk of transitioning to “moderate,” “high,” or “very high” cerebral Aβ-amyloid burden status over 126 months compared to not transitioning (β = −0.005, p = 0.035). Graphical representation of the change in cerebral Aβ-amyloid burden over 126 months, across coffee intake tertiles, demonstrates higher overall rates of Aβ-amyloid accumulation in those participants with “low” coffee intake (Figure 1). Of note, there were no significant differences in baseline Aβ-amyloid across the tertiles of coffee intake.

There were no associations observed between coffee intake and atrophy in gray matter, white matter, or hippocampal volume over 126 months.

The aim of this study was to investigate the relationship between habitual coffee intake and decline in multiple cognitive domains, over 126 months, in 227 cognitively normal individuals, accompanied by examination, in a subset of these individuals, of whether coffee intake was associated with rates of cerebral Aβ-amyloid accumulation, or brain volume atrophy.

Our results showed higher coffee consumption was associated with slower cognitive decline, specifically in the executive function and attention domains. Furthermore, higher consumption of coffee was also associated with slower decline in the AIBL PACC, which has previously been shown to reliably measure the first signs of cognitive decline in at-risk cognitively normal populations (Donohue et al., 2014). Consistent with these findings, higher coffee consumption was additionally associated with lower risk of transitioning from cognitively normal to MCI or AD status over the same time-period of 126 months.

In the neuroimaging subset, higher coffee intake was associated with slower cerebral Aβ-amyloid accumulation over 126 months, as well as lower risk of progressing to “moderate,” “high,” or “very high” Aβ-amyloid burden status over the same time-period. By contrast, coffee consumption was not associated with rates of brain volume atrophy.

The observed effect size estimates suggest that if the average cup of coffee made at home is 240 g, increasing intake from one to two cups per day could potentially provide up to 8% decrease in executive function decline over an 18-month period, and up to 5% decrease in cerebral Aβ-amyloid accumulation over the same time period. However, we readily acknowledge, that further longitudinal observational and intervention studies are required to substantiate this suggestion.

Whilst coffee consumption has previously been linked to decreased risk of MCI and AD (Solfrizzi et al., 2015; Wierzejska, 2017; Wu et al., 2017), until now, no prior studies have characterized the relationship of coffee intake to rates of decline in multiple cognitive domains. Nevertheless, consistent with the longitudinal findings of the current study, in a cross-sectional analysis of community-dwelling Dutch older adults (n = 2914; mean age 59.28 ± 7.2 years), consumption of > 3 cups of coffee per day was associated with better performance on an executive function measure (Araujo et al., 2016). However, the authors reported that this level of coffee consumption was also associated with worse performance on information processing speed and delayed memory measures, compared with 0–1 cups per day. The authors proposed an explanation of their findings whereby the neuro-stimulant effects of coffee on memory function are short-acting, and participants had not consumed coffee for at least 60 min preceding cognitive assessment (Araujo et al., 2016). In contrast to the current report, a study of Finnish twins found coffee consumption was not an independent predictor of cognitive performance in old age, among 2606 middle-aged twins assessed in 1975 and 1981, with a median follow-up of 28 years (Laitala et al., 2009). Cognitive status, however, was assessed using a brief telephone interview comprising a combination of a screen to identify potential dementia cases (TELE) and the Telephone Interview for Cognitive Status (TICS). Furthermore, coffee intake was assessed by the question “How many cups of coffee do you drink daily?” with respondents advised to answer “zero” if they did not drink coffee daily. In the present study, coffee intake over the previous year was used to generate a daily intake value on a continuous scale, such that drinking coffee less frequently than once a day would still yield an intake level; thereby generating a more realistic coffee consumption variable for analyses. Similar to the Finnish study, a study of 3494 Japanese males born between 1900 and 1919 and residing in Hawaii in 1964, reported that coffee intake was not associated with risk of cognitive impairment 25 years later. Like the Finnish study, cognition was not assessed using a comprehensive battery, instead a single screening test, the Cognitive Abilities Screening Instrument (CASI) was utilized, with impairment defined as scoring < 74 out of 100. Moreover, coffee intake was assessed via 24-h dietary recall, and the very specific cohort inclusion criteria limits the generalizability of results (Gelber et al., 2011).

Beyond the current study, only one other published report has examined the relationship between coffee intake and brain Aβ-amyloid (Kim et al., 2019). Consistent with the results of the current longitudinal study, where higher coffee intake was associated with slower cerebral Aβ-amyloid accumulation, this other study, comprising cross-sectional analysis of 411 non-demented older adults from South Korea, found lifetime intake of two or more cups of coffee per day was associated with lower rates of “Aβ positivity” (presence of significant brain Aβ-amyloid), compared to less than two cups per day. However, current coffee intake was not related to “Aβ positivity” in this South Korean cohort (Kim et al., 2019). MRI outcomes were also examined in this cross-sectional study, and consistent with our findings, neither current, nor lifetime coffee intake, was related to cerebral mean cortical thickness or white matter hyperintensity volume (Kim et al., 2019).

Precisely which constituents of coffee contribute to the positive outcomes described in the current study, and by others, remains to be determined. Yet, studies conducted in animal models of AD provide preliminary evidence to suggest the observed benefits are not due to caffeine alone. Both caffeine and crude caffeine (CC), a by-product of the coffee de-caffeination process, partially prevented memory impairment in AD mice, with CC producing a greater effect than pure caffeine. However, only CC consumption reduced the Aβ_1–42 levels and the number of Aβ-amyloid plaques in the hippocampus of these animals. Nevertheless, in vitro, both caffeine and CC were shown to protect primary neurons from Aβ-induced cell death, and to suppress Aβ-induced caspase-3 activity (Chu et al., 2012). Multiple components of coffee, including cafestol and kahweol, have also been show to induce Nrf2 (Nuclear factor-erythroid factor 2-related factor 2) activation in C. elegans (Higgins et al., 2008). Nrf2-dependent upregulation of proteasomal activity has been directly implicated in protection against Aβ-mediated toxicity in cell culture models (Park et al., 2009); however, further work is needed to determine whether this is a related mechanism in humans. Moreover, dietary supplementation with Eicosanoyl-5-hydroxytryptamide (EHT), a minor component of coffee, for 6–12 months in rats, substantially ameliorated cognitive impairment, tau hyperphosphorylation, and elevated levels of cytoplasmic Aβ, potentially through its ability to increase phosphoprotein phosphatase 2A (PP2A) activity. These findings suggest EHT may also make a substantial contribution to the neuroprotective benefits associated with coffee consumption (Basurto-Islas et al., 2014).

The stimulant effect of coffee is attributed to the pharmacological activity of caffeine, acting as an antagonist of adenosine receptors in the brain (de Paulis et al., 2002). Adenosine regulates several physiological functions including sleep, cognitive performance, and memory, and its main role is to regulate neuronal excitatory synaptic transmission by inhibitory A₁ receptors, and synaptic plasticity via facilitatory A₂ receptors (Merighi et al., 2021). The effects of A₁ and A₂ receptor antagonists have been tested in animal models, where inactivation of both A₁ and A₂ receptors was shown to counteract age-related cognitive decline in rats (Prediger et al., 2005) and enhance release of several neurotransmitters including acetylcholine (Boison, 2006). Long-term caffeine administration to amyloid-beta precursor protein (APP) overexpressing mice has been shown to induce significant improvement in multiple cognitive tasks, as well as reduce levels of soluble Aβ-amyloid fragments compared to control mice (Arendash et al., 2006). The average daily intake of caffeine per mouse (1.5 mg) was the human equivalent of 500 mg caffeine, the amount typically found in five cups of coffee per day. Furthermore, caffeine reduced the production of Aβ_1–40 and Aβ_1–42 peptides in neuronal cell cultures from these same APP overexpressing mice (Arendash et al., 2006). Involvement of A₁ receptors has also been observed in APP processing and tau phosphorylation in vitro, and the presence of A₁ receptors in degenerating neurons with neurofibrillary tangles and Aβ-amyloid plaques in the hippocampus and frontal cortex suggest that these receptors may play a role in the pathogenesis of AD (Angulo et al., 2003). Whilst the current study cannot provide evidence that caffeine is causing the positive effects observed, the blocking of adenosine receptors by caffeine, leading to a decrease in Aβ-amyloid in the brain, and subsequent reduction of tau hyperphosphorylation is a potential mechanism that warrants further investigation.

There are some limitations that should be considered when interpreting our findings. Although we adjusted for known potential confounding factors there is a possibility of residual confounding due to other factors which were not measured. There is also the possibility of measurement error or recall bias with respect to the dietary data, which is especially true for questionnaire-based dietary exposures. However, coffee intake is less prone to misreporting due to its long-term habitual nature. Moreover, participants with cognitive impairment at baseline were excluded from the current analysis to reduce the likelihood of measurement error or recall bias. No data on mid-life coffee consumption was obtained from participants, consequently, potential deleterious or beneficial effects of coffee intake at midlife cannot be assessed in the current study. Finally, in our study it was not possible to differentiate between intakes of caffeinated or de-caffeinated coffee, nor to determine the potential consequences of varying methods of coffee preparation (e.g., brewing method, with or without milk or sugar etc.) on the associations observed. Nevertheless, strengths of our study include the long duration of follow-up, and comprehensive characterization of the cohort, including concurrent assessment of both multiple domains of cognition and neuroimaging biomarkers.

In summary, our results support the notion that habitual coffee intake may be a protective factor against AD. Specifically, increased coffee consumption could contribute to slower cognitive decline potentially by slowing the rate of cerebral Aβ-amyloid accumulation, and in doing so, ameliorate the associated neurotoxicity from Aβ-amyloid-mediated oxidative stress and inflammatory processes. Additional longitudinal observational and intervention studies are required to validate our findings and confirm this hypothesis. Such confirmation would corroborate the idea that coffee intake could be incorporated as a modifiable lifestyle factor aimed at delaying AD onset.

The datasets presented in this article are not readily available because the AIBL data are available only to authorized users. Requests to access the datasets should be directed to the following online form: https://ida.loni.usc.edu/collaboration/access/appApply.jsp?project=AIBL.

The study involving human participants was reviewed and approved by the Ethics Commitees of Hollywood Private Hospital, Edith Cowan University, St Vincent’s Health and Austin Health. The patients/participants provided their written informed consent to participate in this study.

SG, SR-S, KT, CM, PM, CR, DA, and RM designed research. SG conducted research, primary responsibility for final content, and analysis for the manuscript. VV, JF, VD, and PB oversaw collection of and analyzed imaging data. SG and SR-S wrote the manuscript. VV, JF, VD, PB, KT, CF, CM, PM, CR, DA, and RM edited the final manuscript. All authors have read and approved the final manuscript.

VV has served as a consultant for IXICO. CM is an advisor to Prana Biotechnology Ltd., and a consultant to Eli Lilly. PM is a full-time employee of Cogstate Ltd. CR has served on scientific advisory boards for Bayer Pharma, Elan Corporation, GE Healthcare, and AstraZeneca, has received speaker honoraria from Bayer Pharma and GE Healthcare, and has received research support from Bayer Pharma, GE Healthcare, Piramal Lifesciences and Avid Radiopharmaceuticals. RM is founder of, and owns stock in, Alzhyme, and is a co-founder of the KaRa Institute of Neurological Diseases. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.