Associations Between Caffeine Consumption, Cognitive Decline, and Dementia: A Systematic Review

Abstract

Background:

Epidemiologic studies have provided inconclusive evidence for a protective effect of caffeine consumption on risk of dementia and cognitive decline.

Objective:

To summarize literature on the association between caffeine and 1) the risk of dementia and/or cognitive decline, and 2) cognitive performance in individuals with mild cognitive impairment (MCI) or dementia, and 3) to examine the effect of study characteristics by categorizing studies based on caffeine source, quantity and other possible confounders.

Methods:

We performed a systematic review of caffeine effects by assessing overall study outcomes; positive, negative or no effect. Our literature search identified 61 eligible studies performed between 1990 and 2020.

Results:

For studies analyzing the association between caffeine and the risk of dementia and/or cognitive decline, 16/57 (28%) studies including a total of 40,707/153,070 (27%) subjects reported positive study outcomes, and 30/57 (53%) studies including 71,219/153,070 (47%) subjects showed positive results that were dependent on study characteristics. Caffeine effects were more often positive when consumed in moderate quantities (100–400 mg/d), consumed in coffee or green tea, and in women. Furthermore, four studies evaluated the relationship between caffeine consumption and cognitive function in cognitively impaired individuals and the majority (3/4 [75% ]) of studies including 272/289 subjects (94%) reported positive outcomes.

Conclusion:

This review suggests that caffeine consumption, especially moderate quantities consumed through coffee or green tea and in women, may reduce the risk of dementia and cognitive decline, and may ameliorate cognitive decline in cognitively impaired individuals.

INTRODUCTION

Dementia is a clinical syndrome characterized by progressive deterioration of cognitive functions and loss of independence in activities of daily living. App-roximately 50 million people are living with dementia worldwide. This number is continuously rising [1], and in 2017 the World Health Organization listed dementia as a public health priority [2]. A range of neuropathological disease entities may underlie a dementia syndrome, including Alzheimer’s disease (AD), vascular pathology (VaD), Lewy bodies (DLB), Parkinson’s disease (PD), or frontotemporal lobar degeneration [1]. Many factors such as cardio- and cerebrovascular disease, metabolism, psychiatric conditions, lifestyle, and education, potentially contribute to the risk of different types of dementia [3]. Furthermore, recent studies have suggested endo- and neurocrine interactions between gut microbiota and the brain (i.e., the microbiota-gut-brain axis [4, 5]) and that dietary factors such as caffeine intake can thereby influence the risk of dementia [6].

Caffeine is a psychoactive substance that is present in many beverages and some foods. The most widely known and consumed caffeine source is coffee, but caffeine can also be found in tea, energy drinks, car-bonated soft drinks, fruits, and cocoa-containing foods [7, 8]. After caffeine ingestion the substance is absorbed into the bloodstream via the gastrointes-tinal tract. From there, caffeine is distributed throughout the entire body. Caffeine biologically acts as an adenosine A₁ and A_2A receptor antagonist, and these receptors are widely distributed throughout the central and peripheral nervous system [9]. By blocking adenosine receptors, caffeine is capable of exerting effects on metabolism, the cardiovascular system, the respiratory system, and neuroinflammatory, neuromodulatory, and neuroprotective processes [10, 11]. More specifically, caffeine may stimulate gastric acid secretion and vasoconstriction, elevate the heart rate and blood pressure, increase the respiratory rate, and ultimately decrease neurodegeneration. Caffeine is able to enhance alertness, wakefulness, psychomotor vigilance, and memory, possibly also through an effect on NMDA receptors [12–14]. Furthermore, caffeine may reduce neuroinflammation and afford neuroprotection, through the consecutive lowering of extracellular calcium, glutamate release from the cell, and microglial activation [15]. There are also health risks associated with excessive caffeine consumption, including anxiety, panic attacks, psychosis, mania, tension, nervousness, irritability, restlessness, nausea, palpitations, insomnia, and diuresis [16].

Research in animal models indicates that caffeine can ameliorate cognitive decline [17]. Studies assessing possible mechanisms underlying this effect have suggested that the effects of caffeine on A_2A receptors can control abnormal synaptic plasticity and synaptotoxicity [18, 19]. Other studies have posited that caffeine intake may delay or reduce the risk of AD by decreasing hippocampal amyloid-β levels in transgenic mice through A_2A receptor blockade [20, 21].

In human epidemiological studies, results for the protective effects of caffeine on cognitive decline and dementia have been mixed. Some studies suggest positive influences of caffeine intake on neurological disorders and dementia [22, 23], while other studies have found no associations between caffeine and de-mentia [24, 25]. The association between caffeine consumption, cognitive decline, and dementia therefore remains inconclusive.

Here, we summarize the available literature on this topic and provide a systematic review. We aimed to address whether there is an association between caffeine and 1) the risk of dementia and/or cognitive decline, and 2) cognitive function in already cognitively impaired individuals (i.e., MCI or dementia). We further aimed to examine the effects of study characteristics (e.g., caffeine source and quantity) and demographic variables of the study sample (e.g., age and sex) on study outcomes.

METHODS

Study selection procedure

We searched the PubMed and Web of Science databases for studies published before June 2, 2020, using the following (combination of) search terms: ‘caffeine’, ‘coffee’, ‘tea’, AND ‘dementia’ OR ‘Al-zheimer(’s)’, AND ‘cognitive’ or ‘cognition’. Only peer-reviewed articles on studies in humans that were published in English, were eligible for inclusion in this pre-determined systematic review. Cross references were additionally assessed for eligibility. We included cognitively unimpaired individuals as well as individuals diagnosed with any type of dementia and/or MCI. The main criteria for article selection were 1) provision of information on the relation between caffeine consumption and the risk of dementia/cognitive decline, and/or 2) assessment of the association of caffeine on cognitive function in individuals with mild cognitive impairment (MCI) or dementia. Because many studies included a mixed sample of persons with dementia and MCI, both groups were taken together and termed ‘cognitively impaired’ subjects. We included any paper that des-cribed original research, regardless of study design, and, therefore, cross-sectional, longitudinal, case-control, controlled trials, cohort, and pilot studies were all assessed in the present review.

Risk of bias assessment

This review was performed according to the preferred reporting items for systematic reviews and meta-analysis (PRISMA) statement (Supplementary Table 1) [26]. The risk of bias for each study was assessed using the Cochrane Collaboration’s tool for non-randomized studies for interventions (ROBINS-I) [27]. Several risk of bias domains were evaluated for each study, including bias due to confounding fac-tors, subject selection, classification of intervention, deviation from intended intervention, missing data, outcome measurement and reporting of results. Each domain was rated as ‘low’, ‘moderate’, ‘serious’, or ‘critical’ risk of bias. An overall risk of bias was de-rived from the quality assessment across all domains of the remaining studies. These judgements were performed independently by two authors (A.C. and C.G.) and final assessment was determined by consensus. Our analyses were confined to studies with low and moderate risk of bias, as studies with serious or critical risk of bias were excluded from the analyses.

Data analysis

Relevant data from the included studies were ex-tracted in piloted forms. Outcome measures in the primary examination were based on overall study outcomes regarding the association between caffeine and 1) the risk of dementia and/or cognitive decline and 2) cognitive function in cognitively impaired in-dividuals. Secondary analyses included examination of the effects of caffeine source (coffee, tea, pure caf-feine, or multiple caffeine containing sources), and quantity (frequency and dosage), and possible confounders (e.g., age or sex), on study outcomes. Based on a previous study [28], the quantity of caffeine con-sumption was divided into three categories: low- (<100 mg/d), moderate- (100–400 mg/d), and high caffeine consumption (>400 mg/d). In accordance with the concentrations of caffeine across sources (i.e., 71–220 mg caffeine/150 ml for coffee and 32–42 mg caffeine/150 ml for tea [29]), moderate caffeine consumption will be defined as 1–4 cups of coffee or 3–10 cups of tea per day. Low caffeine consumption will be defined as <1 cup of coffee or <3 cups of tea per day, and high caffeine consumption will be defined as >4 cups of coffee or >10 cups of tea per day. Outcomes were defined as positive (caffeine improved cognition or slowed down cognitive decline), negative (negative association with cognition), or no association (no relation between caffeine and cognition). Study outcomes could also be mixed, for instance when positive effects were only found in a subset of the sample or when study outcomes were dependent on study characteristics, like caffeine source used.

RESULTS

Study selection and characteristics

The identification of relevant studies is illust-rated in a flow diagram (Fig. 1). Through database searches on PubMed, Web of Science, and cross references, we identified a total of 629 records. First, we excluded 522 articles, including 160 duplicates, based on review of the title and abstract. After full-text assessment of the remaining 107 articles, we excluded 44 articles that had highly overlapping study populations (n = 7), incompatible study designs (n = 4), no suitable cognitive outcome measures (n = 12), only non-caffeine effects (n = 11), or combined interventions (n = 10) (Supplementary Table 2). The remaining 63 studies were assessed for risk of bias, which resulted in the exclusion of two studies [30, 31] (see “Risk of bias” section). The final selection (61 studies) comprised 48 cohort studies, nine case-control studies, three randomized controlled trials, and one pilot study.

Fig. 1.

Flow diagram of identification of relevant studies.

The included studies were published between 1990 and 2020, and were executed in 24 different countries (Table 1): United States of America (n = 10) [24, 25, 32–39], Japan (n = 9) [40–48], China (n = 8) [49–56], United Kingdom (n = 4) [57–60], Finland (n = 3) [28, 61, 62], The Netherlands (n = 4) [62–65], Taipei (n = 3) [66–68], Canada (n = 2) [69, 70], France (n = 2) [71, 72], Portugal (n = 2) [73, 74], Singapore (n = 2) [75, 76], Italy (n = 2) [62, 77], Australia (n = 1) [78], Brazil (n = 1) [79], Germany (n = 1) [80], Iran (n = 1) [81], Ireland (n = 1) [82], Jordan (n = 1) [83], Norway (n = 1) [84], Scotland (n = 1) [85], South Korea (n = 1) [22], Spain (n = 1) [23], Sweden (n = 1) [86], and Switzerland (n = 1) [87]. One study [62] was performed in a multi-national collaboration between Finland, Italy, and the Netherlands. The final selection of articles comprised a total of 153,359 subjects (excluding subjects in the control group), which were either cognitively impaired (AD, DLB, PD, VaD, MCI, or undefined dementia) or cognitively unimpaired.

Table 1.

Characteristics of studies included in the review (n = 61)

Study	Study design retrospective/prospective study length of follow-up	Cohort	Subjects (N and population)	Control (N and population)	Selected cognition measure/domain	Age (y)	Sex (% men)	Caffeine	Effect and principle findings (Positive effect+, negative effect -, no effect/) (HR, OR or RR (95% CI), or p-value)
1. The association between caffeine and the risk of dementia and/or cognitive decline
1. Al-khateeb et al.	Cross-sectional	Senior homes	52 dementia	50 cognitively	MMSE	69.8	61%	Coffee	+, protective effect
2014 [83] Jordan	case-control	and Jordan		healthy		(7.4)		retrospective	of coffee against
	study	University							cognitive decline,
	retrospective	Hospital							with a 6.25-fold lower
	NA								risk with increased intake.
									OR = 0.16 (0.066–0.37)
2. Arab et al.	Longitudinal	The Cardiovascular	X/4,809 subjects,	X/4,809 subjects,	3MSE	72.6	43%	Coffee,	+, association coffee
2011 [32] USA	cohort	Health	caffeine consumer	non-caffeine		(5.4)		Tea (NS)	and tea consumption
	study	Study (CHS)		consumer				retrospective	with reduced rates
	prospective								of cognitive decline
	median:								in women. Tea: p = 0.007
	7.9 y								Coffee: p = 0.02 /,
									no association in men.
									Tea: p = 0.67
									Coffee: p = 0.99
3. Araújo et al.	Cross-sectional	The Longitudinal	13,165 subjects,	1,398 subjects,	Learning,	52.0	46%	Coffee	+, moderate coffee
2015 [79] Brazil	cohort study	Study of Adult	coffee consumer	non-/low coffee	recall, and	(9.0)		retrospective	consumption (2–3 cups/d)
	retrospective	Health (ELSA-Brasil)		consumer	word				associated with better
	12 mo				recognition				cognitive function in
					tests				elderly only (65–74 y).
									p = 0.025 /, no association
									for low and high coffee
									consumption, and
									for adults (35–64 y).
4. Araújo et al. 2016	Longitudinal	The Rotterdam	2,914 subjects	NA	LDST	59.3	45%	Coffee	+, higher coffee
[63] The Netherlands	and cross-	Study 2005–2009				(7.2)		retrospective	consumption (≥3 cups/d)
	sectional								associated with better
	cohort study								cognitive performance.
	prospective								p = 0.026 /, no association
	5.5 y								in a longitudinal model.
5. Beydoun et al.	Longitudinal and	The Baltimore	3,047 subjects,	3,047 subjects,	MMSE	58.9	60%	Multiple sources	+, association caffeine
2014 [33] USA	cross-sectional	Longitudinal	follow-up	baseline		(18.0)		(NS)	intake with better global
	cohort study	Study of						prospective	cognitive function at
	prospective	Aging (BLSA)							baseline for age >70 y.
	46 y								p = 0.008 /, no association
									found for age <70 y.
6. Boot et al.	Cross-sectional	The Mayo Clinic	383 cognitively	294 cognitively	NA	82.4	63%	Multiple sources	+, association caffeine
2013 [34] USA	case-control	Study of Aging;	impaired	healthy		(7.5)		(coffee, tea and	intake and reduced risk
	study retrospective	The Alzheimer	(236 AD, 147 DLB)					caffeinated soda)	of DLB. OR = 0.29
	NA	Disease Patient							(0.14–0.57) retrospective
		Registry; Alzheimer
		Disease Research
		Center Study
7. Broe et al.	Cross-sectional	The Repatriation	170 AD	170 cognitively	MMSE	78.1	38%	Coffee,	/, no association between tea
1990 [78] Australia	case-control	General Hospital		healthy		(7.3)		Tea (NS)	and coffee consumption and
	study retrospective	Concord and						retrospective	reduced risk of AD. Tea:
	NA	Lidcombe Hospital							OR = 1.42 (0.93–2.17) Coffee:
									OR = 2.25 (0.72–7.71)
8. Chen et al.	Longitudinal	The Chinese	1306	4385	MMSE	82.9	24%	Tea (NS)	+, association tea drinking
2012 [49] China	case-control	Longitudinal Health	cognitive decline	cognitive healthy	(Chinese	(11.0)		retrospective	with cognitive decline.
	study	Longevity Study			version)				p = 0.0468
	prospective	(CLHLS) 2002
	3 y
9. Chin et al.	Cross-sectional	The Dublin Healthy	466 cognitively	NA	MMSE	75.5	45%	Tea (NS)	+, tea intake positively
2008 [82] Ireland	cohort study	Ageing Study	healthy			(6.1)		retrospective	correlated with global
	retrospective NA								cognitive performance.
									p = 0.042
10. Chuang et al.	Longitudinal and	The Nutrition and	516 subjects,	912 subjects,	SPMSQs	73.6	51%	Coffee,	+, higher intake (moderate
2019 [66] Taipei	cross-sectional	Health Survey	caffeine	non-/low caffeine	and MMSE	(0.8)		Tea (NS)	consumption; ≥7 times/wk)
	cohort study	in Taiwan	consumer	consumer	(Chinese			retrospective	of tea and coffee associated
	prospective	(NAHSIT) 2014–2016						version)	with lower risk of dementia
	11 y	and 1999–2000							Coffee: OR = 0.55 (0.30–0.98)
									Tea: 0.46 (0.28–0.78) /,
									no association with low
									coffee and tea consumption,
									and in men only.
11. Corley et al.	Cross-sectional,	The Lothian Birth	893 subjects	NA	Memory	69.5	48%	Multiple sources	+, general cognitive ability
2010 [85] Scotland	cohort study	Cohort 1936 Study;				(0.8)		(14 caffeine-	and memory with adjustments
	retrospective	The Scottish						containing items,	for age and sex. p = 0.02
	2–3 mo	Mental Survey 1947						e.g., coffee, tea,	/, no association with cognitive
								chocolate, etc.)	function when additionally
								retrospective	adjusted for occupational
									social class and
									childhood IQ. p = 0.11
12. Dai et al.	Longitudinal	The KAME Project	1,275 subjects,	315 subjects,	CASI	71.8	46%	Tea (NS)	/, no association tea intake
2006 [36] USA	cohort study		tea consumer	non-/low tea		(NA)		retrospective	and risk of incident probable AD.
	prospective			consumer					HR = 1.29 (0.63–2.64)
	6.3 y
13. Dong et al.	Cross-sectional	National Health and	1,803 subjects,	710 subjects,	DSST	NA;	48%	Coffee	+, association between moderate
2020a [50] China	cohort study	Nutrition Examination	coffee consumer	non-coffee		>60 y		retrospective	and high coffee consumption
	retrospective	Survey (NHANES)		consumer					and cognitive performance.
	24 h	2011–2014							Moderate: OR = 0.71 (0.47–0.87)
									High: OR = 0.56 (0.39–0.79)
									/, no association with low
									and high coffee consumption.
									Low: OR = 1.36 (0.96–1.93)
14. Driscoll et al.	Longitudinal and	Women’s Health	2,541 subjects,	2,926 subjects,	3MSE	Range:	0%	Multiple sources	+, association between caffeine
2016 [37] USA	cross-sectional	Initiative	caffeine consumer	non-low caffeine		[65–80]		(coffee, tea and	intake and probable dementia.
	cohort study	Memory Study		consumer				cola) retrospective	HR = 0.74 (0.56–0.98)
	prospective								+, stronger association with
	10 y								higher caffeine intake over
									time. p < 0.0001
15. Eskelinen et al.	Longitudinal	The Cardiovascular	X/1,409 subjects,	X/1,409 subjects,	MMSE	71.3	38%	Coffee,	+, lower risk of dementia and AD
2009 [28] Finland	cohort study	Risk Factors, Aging	caffeine consumer	non-caffeine		(4.0)		Tea (NS)	for moderate (3–5 cups) coffee
	prospective	and Dementia study		consumer				retrospective	consumption. Coffee (moderate):
	21 y	(CAIDE) (North							OR = 0.34 (0.16–0.73),
		Karelia Project							/, no association found for tea
		and FINMONICA study)							consumption and high
									(>5 cups/d) coffee consumption.
									Coffee (high): OR = 0.61
									(0.30–1.21) Tea:
									OR = 1.04 (0.59–1.84)
16. Feng et al.	Longitudinal	The Chinese Longitudinal	3,187 subjects,	3,952 subjects,	Verbal	91.4	42.9%	Tea (NS)	+, association daily tea drinking
2012 [75] Singapore	cohort study	Health Longevity	tea consumer	non-tea consumer	fluency test	(7.5)		retrospective	and better cognitive function.
	prospective	Study (CLHLS) 1998							p = 0.02
	7 y
17. Feng et al.	Longitudinal	The Osteoporotic	1,430 subjects,	2,414 subjects,	3MSE	72.4	100%	Tea	/, no association black tea
2018 [25] USA	cohort study	Fractures in	tea consumer	non-tea consumer		(5.2)		(black tea)	consumption and cognitive
	prospective	Men (MrOS) Cohort						retrospective	decline. OR = 1.19 (0.81–1.75)
	6.8 y
18. Fischer et al.	Longitudinal	Aging, Cognition and	2,622 subjects	2,622 cognitively	CERAD	81.2	35%	Coffee, Tea	/, no association coffee and green
2018 [80] Germany	cohort study	Dementia in Primary	(2,204 cognitively	healthy, baseline	memory score	(3.4)		(green tea)	tea intake with memory/cognitive
	prospective	Care Patients	healthy, 418 incident					retrospective	decline or incident AD. Coffee:
	10 y	(AgeCoDe) Cohort	dementia),						HR = 0.97 (0.90–1.04) Green tea:
									HR = 0.94 (0.86–1.02)
19. Gelber et al.	Longitudinal	The Honolulu-Asia	2,787 cognitively	707 cognitively	CASI	52.5	100%	Coffee,	/, no association between coffee
2011 [24] USA	case-control	Aging Study	healthy, coffee	healthy, non-/		(4.5)		Multiple sources	or general caffeine intake and risk
	study prospective	(HAAS)	consumer	low coffee				(coffee, tea, cola)	of dementia and cognitive impairment.
	25 y			consumer				retrospective	OR = 1.05 (0.58–1.90)
20. Gu et al.	Cross-sectional	The Weitang	1,570 subjects	3,008 subjects (2,218	AMT	67.6	48%	Tea (green and	+, inverse association between
2018 [51] China	case-control	Geratric Diseases	(1,416 cognitively	cognitively healthy,		(6.3)		other tea types)	habitual and (green) tea consumption
	study	Study	healthy, 155	790 cognitively				retrospective	(>5 times/wk) and prevalence
	retrospective		cognitively	impaired),					of cognitive impairment.
	NA		impaired),	non-habitual					OR = 0.74 (0.56–0.99)
			habitual tea	tea consumers					/, no association with green tea
			consumers						consumption of 1–5
									times/wk and other tea types.
									1–5 times/wk: OR = 0.56 (0.29–1.07)
									Other tea: OR = 0.66 (0.37–1.18)
21. Haller et al.	Longitudinal	Elderly in Geneva	145 subjects,	145 subjects,	MMSE	73.8	44%	Coffee	+, association moderate coffee
2018 [87] Switzerland	cohort study	and Lausanne	follow-up	baseline		(3.5)		retrospective	consumption and reduced risk
prospective	counties	of deteriorated cognition (dCON).
	3 y								OR = 0.45 (0.21–0.95) /, no association
									for low coffee consumption.
22. Huang et al.	Cross-sectional	The Project of	429 cognitively	252 cognitively	MMSE	93.5	33%	Tea (NS)	+, tea consumption associated with
2009 [52] China	cohort study	Longevity and Aging	impaired	healthy		(3.3)		retrospective	cognitive impairment in men.
	retrospective	in Dujiangyan (PLAD)							/, no association in women.
	2 y
23. Iranpour et al.	Cross-sectional	National Health	1,065 subjects,	375 subjects,	DSST	69.8	51%	Multiple sources	+, positive association between high
2020a [81] Iran	cohort study	and Nutritional	≥Q2 caffeine	Q1 caffeine		(2.3)		(e.g., tea, soda,	caffeine intake and cognitive
	retrospective	Examination Surveys	consumer	consumer				chocolate, etc.)	function in an univariate model.
	24 h	(NHANES) 2013–2014	retrospective						p = 0.004 /, no association with
									multiple adjustments. p = 0.99
24. Jarvis	Cross-sectional	The Health and	X/7,414 subjects,	X/7,414 subjects,	Reaction time,	NA; ∼45	45%	Coffee,	+, association increased levels
1993 [57] UK	cohort study	Lifestyle Survey	caffeine consumer	non-caffeine	incidental verbal			Tea (NS)	of coffee and tea consumption
	retrospective			consumer	memory and visuo-			retrospective	with improved cognitive performance.
	NA				spatial reasoning				Stronger association for coffee
									than tea intake, and
									older people than younger
									people. p < 0.05
25. Johnson-Kozlow	Cross-sectional	The Rancho	1,528 cognitively	NA	MMSE	72.9	42%	Coffee	+, association between higher
et al. 2002 [38] USA	cohort study	Bernardo Study,	healthy			(9.0)		retrospective	lifetime coffee consumption
	prospective	1988–1992							and better cognitive performance
	NA								in women. p = 0.023
									/, no association in men.
26. Kitamura et al.	Cross-sectional	The Project in	601 subjects	539 subjects	MMSE	68.9	55%	Tea	+, green tea intake associated with
2016 [42] Japan	cohort study	Sado for Total	(490 cognitively	(406 cognitively		(10.6)		(green tea)	/, no association in men.
	retrospective	Health (PROST)	healthy, 111	healthy, 133				retrospective	OR = 0.73 (0.54–0.99)
	NA	2008–2014	cognitively	cognitively
			impaired),	impaired), non-
			tea consumer	tea consumer
27. Konishi et al.	Cross-sectional	Healthy Japanese	50 cognitively	50 cognitively	SAT: executive	Range:	50%	Pure caffeine	+, better executive function with
2018 [48] Japan	RCT prospective	volunteers, 2016	healthy, caffeine	healthy,	function	[22–59]	prospective	caffeine consumption. p = 0.03
	30 min		consumer	Placebo
28. Kuriyama et al.	Cross-sectional	The Tsurugaya	833 subjects,	170 subjects,	MMSE	74.7	43%	Coffee,	+, green tea consumption of >2
2006 [43] Japan	cohort study	Project	caffeine consumer	non-/low caffeine	(Japanese	(4.6)		Tea (green,	cups/d and prevalence of cognitive
	retrospective			consumer	version)			black/oolong	impairment. OR = 0.46 (0.30–0.72)
	NA							tea)	/, green tea consumption of 4–7 cups/wk.
								retrospective	OR = 0.62 (0.33–1.19) /, no association
									for coffee and black/oolong tea.
									Coffee: OR = 1.03 (0.59–1.80)
									Black/oolong: OR = 0.87 (0.55–1.38)
29. Laitala et al.	Longitudinal	Finnish Twin	2606 cognitively	NA	TELE	74.4	52%	Coffee	/, no association between coffee and
2009 [61] Finland	cohort study	Cohort Study	healthy			(5.3)		retrospective	cognitive performance.
	prospective								OR = 1.07 (0.97–1.17)
	Median:
	28 y
30. Larsson &Wolk	Longitudinal and	The National Research	X/28,775 subjects,	X/28,775 subjects,	NA	83.2	53%	Coffee	/, coffee consumption not associated
2018 [86] Sweden	cross-sectional	Infrastructure SIMPLER	caffeine consumer	non-/low caffeine		(5.1)		retrospective	with risk of AD. HR = 1.01 (0.86–1.18)
	cohort study	(Swedish Infrastructure		consumer
	prospective	for Medical Population-
	12.6 y	based Life-course
		Environmental Research);
		Swedish Mammography
		Cohort and the
		Cohort of Swedish Men
31. Lee et al.	Cross-sectional	A Nationwide Survey	X/7,964 subjects,	X/7,964 subjects,	TMSE	75.7	50%	Coffee,	+, inverse associations with dementia.
2017 [67] Taipei	cohort study	in Japan, 2011–2013	caffeine consumer	non-caffeine		(6.6)		Tea (green and	Coffee: OR = 0.59 (0.35–0.97)
	retrospective			consumer				other tea types)	Green tea: OR = 0.51 (0.34–0.75)
	NA							retrospective	Other tea: OR = 0.41 (0.28–0.60)
32. Lesk et al.	Cross-sectional	The Oxford Project to	57 subjects,	32 subjects,	MMSE	Range:	38%	Multiple sources	/, no association between caffeine-
2009 [58] UK	cohort study	Investigate Memory	caffeine consumer	non-caffeine		[67–95]		(e.g., coffee, tea,	containing foodstuffs (CCFS)
	retrospective	and Ageing		consumer				soft drinks,	and cognitive decline.
	4 h	(OPTIMA) cohort						chocolate, etc.)
								retrospective
33. Lindsay	Longitudinal	The Canadian Study	194 AD	3,894 cognitively	3MSE	Range:	42%	Coffee,	+, regular (nearly every day) coffee
2002b [69] Canada	case-control	of Health and	healthy	[69–105]	Tea (NS)	consumption and a reduced risk of AD.
	study prospective	Aging (CSHA)						retrospective	OR = 0.69 (0.50–0.96) /, no association
	5 y								with tea drinking. OR = 1.12 (0.78–1.61)
34. Maia &	Cross-sectional,	Dementia Outpatient	54 AD	54 cognitively	MMSE	70.8	48%	Multiple sources	+, lower risk for AD, independent
de Mendonça 2002	case-control	Clinics, Hospital		healthy		(7.7)		(e.g., coffee, tea,	of confounding variables.
[73] Portugal	study	Santa Maria, Lisbon						cola, etc.)	OR = 0.40 (0.25–0.67)
	retrospective							retrospective
	20 y
35. Mirza et al.	Longitudinal	The Rotterdam	3,876 subjects,	492 subjects,	MMSE	NA; ∼70	41%	Coffee	+, association between coffee consumption
2014 [64]	cohort study	Study 1989–1990	coffee consumer	non-/low coffee				retrospective	(>3 cups/d) and incident dementia with
The Netherlands	prospective			consumer					short (0–4 y) follow-up. Short-term:
	8.7 y								HR = 0.70 (0.51–0.96) -, increased risk
									of incident dementia for
									long-term effect (>4 y),
									possibly due to reverse
									causality. Long-term:
									HR = 1.14 (0.83–1.56)
36. Ng et al.	Longitudinal and	The Singapore	X/2,501 subjects,	X/2,501 subjects,	MMSE	66.0	36%	Coffee, Tea	+, association regular black/oolong
2008 [76] Singapore	cross-sectional	Longitudinal	caffeine consumer	non-caffeine		(7.7)		(green and black/	and green tea consumption with lower
	cohort study	Ageing Studies		consumer				oolong tea)	prevalence of cognitive impairment,
	prospective	(SLAS) cohort						retrospective	and black/oolong tea with reduced risk
	median:16 mo								of cognitive decline over time. Black/oolong:
									OR = 0.55 (0.40–0.76) Green tea:
									OR = 0.42 (0.25–0.69) /, no association
									for coffee. Coffee: OR = 0.99 (0.69–1.45)
37. Noguchi-Shinohara	Longitudinal	The Nakajima	X/490 subjects,	X/490 subjects,	MMSE	71.2	33%	Coffee, Tea	+, reduced risk of cognitive decline
et al. 2014 [44]Japan	cohort study	Project	caffeine consumer	non-caffeine		(6.4)		(black and green tea)	with green tea. Green tea:
	prospective			consumers				retrospective	OR = 0.53 (0.30–0.93)
	4.9 y								/, no effect for coffee or black tea
									on incidence of dementia or MCI.
									Coffee: OR = 1.22 (0.63–2.36)
									Black tea: OR = 1.19 (0.64–2.24)
38. Nurk et al.	Cross-sectional	The Hordaland	1,083 subjects,	948 subjects,	modified	Range:	45%	Tea (NS)	+, habitual tea intake associated with
2009 [84] Norway	cohort study	Health Study	tea consumer	non-tea consumer	MMSE	[70–74]		retrospective	better cognitive test performance.
	retrospective	(HUSK),							p = 0.046
	NA	Norway
39. Paganini-Hill	Longitudinal	The 90 + Study,	587 subjects	587	MMSE/	93	NA	Multiple sources	+, caffeine consumption of >200 mg/d
et al. 2016 [39] USA	cohort study	The Leisure World	(268 incident	cognitively healthy	CASI	(2.6)		(e.g., coffee, black	associated with reduced risk of dementia
	prospective	Cohort Study	dementia),	elderly, baseline				tea, green tea, soft	compared with caffeine consumption
	36 mo		follow-up					drinks, chocolate,	of <50 mg/d at age of 90. HR = 0.66
								etc.) retrospective	(0.43–0.99) /, no association found
									20 y earlier, at age 70, or lower
									caffeine consumption at age 90.
40. Ritchie et al.	Longitudinal	The Three-City	7,017 cognitively	7,017 cognitively	Isaacs	73.7	40%	Multiple sources	+, inverse association between coffee
2007 [71] France	cohort study	Study (Bordeaux,	healthy, follow-up	healthy, baseline		(5.2)		(tea and coffee)	consumption (>300 mg/day)
	prospective	Dijon, Montpellier)						retrospective	and cognitive decline in women,
	3.5 y								especially at higher ages.
									OR = 0.67 (0.53–0.85)
									/, no effect found in men.
									OR = 1.18 (0.87–1.59)
41. Santos et al.	Longitudinal	Elderly in Porto	309 cognitively	309 cognitively	MMSE	70	41%	Multiple sources	+, caffeine intake (>62 mg/day)
2010 [74] Portugal	cohort study	healthy, follow-up	healthy, baseline		(1.9)		(82 caffeine-containing	associated with lower risk of
	prospective							food items, e.g., coffee,	cognitive decline in women.
	median: 48 mo							tea, soft drinks, chocolate,	RR = 0.51 (0.27–0.97)
								etc.) retrospective	/, no effect found in men. RR = 0.51 (0.22–1.16)
42. Shen et al.	Cross-sectional	The Zhejiang Major	2,530 subjects,	6,845 subjects,	MMSE	70.0	48.5%	Tea	+, for black tea consumption
2015 [53] China	cohort study	Public Health	caffeine consumer	non-caffeine	(Chinese	(7.7)		(black, green and	and prevalence of cognitive
	retrospective	Surveillance		consumer	version)			multiple tea types)	impairment. Black tea:
	NA	Program (ZPHS) 2014						retrospective	OR = 0.48 (0.29–0.80)
									+, positive association for 2–4 cups/d
									and > 4 cups/d tea consumption in general.
									2–4 cups/d: OR = 0.71 (0.59–0.84)
									≥4 cups/d: OR = 0.76 (0.63–0.91)
									/, no association for green
									tea and low tea consumption
									in general. Green tea:
									OR = 1.00 (0.74–1.35)
									<2 cups/d: OR = 1.09 (0.88–1.35)
43. Shirai et al.	Longitudinal	The National Institute	X/1,305 subjects,	X/1,305 subjects,	MMSE	66.7	48%	Coffee,	+, association between 2–3 times/d and
2020 [45] Japan	cohort study	for Longevity Sciences,	caffeine consumer	non-caffeine	(Japanese	(6.2)		Tea (green tea)	>4 times/d green tea consumption and
	prospective	Longitudinal Study		consumer	version)			retrospective	reduced risk of cognitive decline.
	5.3 y	of Aging (NILS-LSA)							2–3 times/d: HR = 0.71 (0.52–0.97)
									≥4 times/d: HR = 0.72 (0.54–0.98)
									/, no association for <once/d green
									tea consumption and general
									coffee consumption in general.
44. Smith	Cross-sectional	The Bristol Stress and	X/3,223 subjects,	X/3,223 subjects,	CFQ	49.6	43%	Multiple sources	+, caffeine consumption reduces
2009 [59] UK	cohort study	Health at Work Study &	caffeine consumer	non-caffeine		(21.9)		(caffeinated drinks)	risk on cognitive failures.
	respective	The Cardiff Health and		consumer				retrospective	p < 0.005
	NA	Safety and Work Study
45. Solfrizzi et al.	Longitudinal	The Italian Longitudinal	985 subjects,	460 subjects,	MMSE	71.8	56%	Coffee	+, lower rate of MCI incidence for
2015 [77] Italy	cohort study	Study on Aging (ILSA)	coffee consumer	non-/low coffee		(5.0)		retrospective	moderate (1–2 cups/d) coffee consumption.
	prospective			consumer					HR = 0.31 (0.13–0.75) /, no association
	median:								with low coffee consumption (<1 cup/d).
	3.5 y								HR = 0.47 (0.211–1.02) -, higher rate
									of incidence MCI for change in coffee
									consumption habits.
									HR = 1.80 (1.11–2.92)
46. Sugiyama et al.	Longitudinal	The Ohsaki Cohort 2006	11,089 subjects,	2,048 subjects,	Dementia	73.6	45%	Coffee	+, coffee consumption associated
2016a [46] Japan	cohort study		caffeine consumer	non-caffeine	Scale	(5.8)		retrospective	with lower risk of incident dementia.
	prospective			consumer					HR = 0.72 (0.61–0.84)
	5.7 y
47. Tomata et al.	Longitudinal	The Ohsaki Cohort 2006	11,411 subjects,	2,234 subjects,	CDR	73.8	44%	Tea (green, black	+, Green tea consumption of >3 cups/d
2016a [47] Japan	cohort study	caffeine consumer	non-/low caffeine		(5.8)		and oolong tea)	associated with a lower risk of incident
	prospective			consumer				retrospective	dementia. Green tea: OR = 0.74 (0.63–0.88)
	5.7 y								/, no association found for low green tea
									consumption (<2 cups/d) and black
									and oolong tea. Green tea: OR = 0.94
									(0.79–1.11) Black tea:
									HR = 0.98 (0.61–1.59)
									Oolong tea: HR = 0.89 (0.54–1.45)
48. Tyas et al.	Longitudinal	The Manitoba Study	36 AD	658 cognitively	3MSE	74.0	37.6%	Coffee,	/, no association between coffee and
2001b [70] Canada	cohort study	of Health and Aging		healthy		(5.8)		Tea (NS)	tea consumption and risk of AD.
	prospective	(MSHA); The Canadian						retrospective	Coffee: RR = 1.03 (0.47–2.30)
	5 y	Study of Health							Tea: RR = 0.46 (0.20–1.06)
									and Aging (CSHA)
49. Valls-Pedret et al.	Cross-sectional,	The Prevención	447 cognitively	NA	RAVLT –	Range:	48%	Coffee	+, better memory function and global
2012 [23] Spain	cohort study	con Dieta	healthy		delayed recall	[55–80]		retrospective	cognition with coffee consumption.
	retrospective	Mediterránea							Coffee: p = 0.016
	NA	(PREDIMED) Study
50. Van Boxtel et al.	Longitudinal	The Maastricht	1,366 cognitively	1,366 cognitively	VVLT	50.2	52%	Multiple sources	/, no association found between
2003 [65]	cohort study	Aging Study (MAAS)	healthy, follow-up	healthy, baseline		(15.4)		(coffee, tea, cola,	caffeine consumption
The Netherlands	prospective							energy-drink)	and age over time.
	6 y							retrospective
51. Van Gelder et al.	Longitudinal	The Finland, Italy	531 cognitively	145 cognitively	MMSE	76.1	100%	Coffee	+, inverse association between low
2007 [62] Finland,	cohort study	and the Netherlands	healthy	healthy, non-coffee		(4.2)		retrospective	and moderate coffee consumption
Italy,	prospective	(FINE) Study		consumer					(<4 cups/day) and cognitive decline.
The Netherlands	10 y	cohorts							3 cups/d:
									p < 0.001 1, 2, 4 cups/d: p < 0.05
									/, no effect for high coffee consumption
									(>4 cups/d). p > 0.05
52. Vercambre et al.	Longitudinal	The Women’s	X/2,475 cognitively	X/2,475 cognitively	Global	NA;	0%	Multiple sources	+, slower rates of global cognitive decline
2013 [72] France	cohort study	Antioxidant	healthy,≥Q2	healthy, Q1 caffeine	cognitive	>65 y		(116 caffeine-containing	with increasing caffeine intake
	prospective	Cardiovascular	caffeine	consumer	score			items, e.g., caffeinated	(4 cups/d versus non-/low caffeine
	5 y	Study (WACS)	consumer					coffee, decaffeinated	consumption). p = 0.02 +, stronger
		Cohort						coffee, tea, chocolate)	association with multiple additional
								retrospective	adjustments and additional vitamin B
									supplementation. p = 0.02
									/, adjustments only for age,
									education and energy from diet. p = 0.066
53. Walters &Lesk	Cross-sectional	Division of Psychology,	20 cognitively	20 cognitively	MMSE	73.4	NA	Pure caffeine	/, no significant interaction
2016 [60] UK	RCT	University of	healthy, caffeine	healthy, placebo		(6.6)		prospective	of caffeine found on cognitive tests.
	prospective	Bradford database	consumer
	NA
54. Wang et al.	Longitudinal and	Elderly in Shanghai	224 MCI	781 cognitively	MMSE	72.7	42%	Tea (NS)	+, tea can protect people >60
2017 [54] China	cross-sectional	(from Huangpu,		healthy		(8.5)		retrospective	y against MCI.
	cohort study	Changning, Putuo,							OR = 0.59 (0.43–0.82)
	prospective	Pudong districts)							/, tea consumption in the age >70 y.
	1 y								OR = 0.72 (0.49–1.07)
55. Wu et al.	Cross-sectional	The National Health	X/2,219 subjects,	X/2,219 subjects,	MMSE	73.3	52%	Coffee,	+, decreased risk of cognitive
2011 [68] Taipei	cohort study	Interview Survey 2005	caffeine consumer	non-caffeine		(5.9)		Tea (NS)	impairment with coffee.
	retrospective			consumer				retrospective	Coffee: OR = 0.51 (0.31–0.83)
	NA								/, no effect for tea. Tea:
									OR = 0.99 (0.75–1.30)
56. Xu et al.	Cross-sectional	China Longitudinal	439 MCI	1,692 cognitively	MMSE	70.9	45%	Tea	+, protective effect against MCI
2018 [55] China	cohort study	Aging Study (CLAS)	healthy	(7.9)	(green, black,	for green tea consumption in men,
	retrospective							oolong tea)	particularly at <70 y. Green tea
	NA							retrospective	(age <70): OR = 0.376 (0.20–0.70)
									/, green tea in women; black and
									oolong tea in general.
									Green tea (women): OR = 0.82
									(0.58–1.16) Black tea:
									OR = 0.74 (0.37–1.49)
57. Yang et al.	Cross-sectional	Elderly in	847 subjects	11,68 subjects	MMSE	79.5	42%	Tea (NS)	+, association between tea consumption
2016 [56] China	cohort study	Zhejiang province	(749 cognitively	(822 cognitively		(7.6)		retrospective	and AD or severe cognitive impairment.
	retrospective		healthy +98	healthy+346					OR = 0.5 (0.4–0.6)
	NA		dementia), tea	dementia),
			consumer	non-tea consumer
2. The association between caffeine and cognitive function in cognitively impaired individuals
58. Cao et al.	Longitudinal	Florida Alzheimer’s	124 subjects	124 subjects	MMSE	74.9	40%	Multiple sources	+, caffeine/coffee intake associated
2012 [35] USA	case-control	Disease Research	(69 cognitively	(69 cognitively		(1.9)		(Plasma caffeine	with reduced risk of dementia
	study	Center (FADRC),	healthy, 32 MCI,	healthy, 32 MCI,				concentration)	or delayed onset,
	prospective	Miami and	23 dementia),	23 dementia),				retrospective	particularly for those who
	2–4 y	Tampa cohort	follow-up	baseline					already have MCI. p < 0.02
59. Cho et al.	Cross-sectional,	The Movement	136 PD, coffee	60 PD,	K-MMSE	66.3	52%	Coffee	+, better global cognitive scores
2018 [22]	cohort study	Disorders Clinic,	consumer	non-coffee	(Korean	(9.5)		retrospective	for coffee consumption
South Korea	retrospective	Chonnam National		consumer	version)				in patients with PD.
	NA	University Hospital							p = 0.004
60. Ide et al.	Longitudinal	The White Cross	12 cognitively	12 cognitively	MMSE-J	88	17%	Green tea	+, association between three-month
2014 [40] Japan	pilot study	Nursing Home in	impaired (3 AD,	impaired (3 AD,	(Japanese	(7.6)		powder	green tea consumption and improved
	prospective	Higashi-Murayama,	8 VaD,	8 VaD,	version)			prospective	cognitive function or reduced
	3 months	Tokyo, Japan 2012	1 DLB),	1 DLB),					progression of cognitive
			follow-up	baseline					dysfunction.s p = 0.03
61. Ide et al.	Longitudinal	The White Cross	17 cognitively	16 cognitively	MMSE-J	84.8	12%	Green tea	/, no association between
2016 [41] Japan	RCT	Nursing Home	impaired (9 AD,	impaired (8 AD,	(Japanese	(9.3)		powder	1-y green tea consumption
	prospective	in Higashi-Murayama,	7 VaD,	8 VaD),	version)			prospective	and cognitive performance.
	12 months	Tokyo, Japan	1 DLB),	placebo					p = 0.59
			caffeine
			consumer

MCI, mild cognitive impairment; AD, Alzheimer’s disease; PD, Parkinson’s disease; DLB, dementia with Lewy bodies, VaD, vascular dementia; RCT, randomized controlled trial; AMT, Abbreviated Mental Test; MMSE, Mini-Mental State Examination; 3MSE, Modified Mini-Mental State Examination; CERAD. Consortium to Establish a Registry for Alzheimer’s Disease; CASI, Cognitive Abilities Screening Instrument; DSST, Digit Symbol Substitution Test; CFQ, Cognitive Failures Questionnaire; MSQ, Mental Status Questionnaire; TELE, Telephone-Assessment of Cognitive State; TMSE, Tested Thai Mental State Examination; CDR, Clinical Dementia Rating; LDST, Letter-Digit Substitution Task; RAVLT, Rey Auditory Verbal Learning Test; VVLT, Visual Verbal Learning Test; SAT, Shifting Attention Test; SPMSQs, Short Portable Mental Status Questionnaires; NA, not available; NS, non-specified; HR, hazard ratio; OR, odds ratio; RR, relative risk; CI, confidence interval; y, year; mo, month; wk, week, d, day; h, hour, min, minute. Age values represent mean (±SD), unless otherwise indicated. ^aOverlapping or sharing population but different study design. ^bSmall number of overlapping population with other included study.

Risk of bias

Using the Cochrane Collaboration tool, an assessment of bias was performed for all included studies, which lead to the exclusion of two studies [30, 31] (Supplementary Table 3). Furthermore, 39/61 studies had low risk of bias and 22/61 studies had moderate risk of bias. Assessment of bias across risk of bias domains revealed predominantly moderate- to low risk of bias for six out of seven domains (Fig. 2). High risk of bias was observed on the ‘deviations from intended interventions’ domain, which could be explained by most studies employing self-reported data.

Fig. 2.

Risk of bias assessment of the included studies.

Associations between caffeine consumption and cognition

Caffeine and the risk of dementia/cognitive decline

Of the 61 articles included in this review, 57 studies with a total of 153,070 subjects, assessed the association between caffeine and the risk of dementia and/or cognitive decline (Fig. 3A, B). Within these studies, 16/57 (28%) studies including 40,707/153,070 (27%) subjects found a positive association for caffeine on the risk of dementia and/or cognitive decline that was independent of study related factors. Approximately half of the studies (30/57 (53%) studies including 71,219/153,070 (47%) subjects) reported positive results that were dependent on caffeine consumption quantity (n = 14), type of caffeine source (n = 11), sex (n = 7), age (n = 4), caffeine consumption duration (short- or long-term effects) (n = 2), and/or adjustments for covariates (n = 3). No association between caffeine and risk of dementia or cognitive decline was found in 11/57 (19%) studies including 41,144/153,070 (27%) subjects.

Fig. 3.

A Study outcomes for the association between caffeine and dementia and/or cognitive function. Pie charts show study outcomes based on population, caffeine consumption dosage and type of caffeine source: positive effect (darker green), positive effect dependent on study characteristics (lighter green), no effect (gray), and negative effect (red [none observed]). Outlined charts indicate a predominant positive outcome.

Fig. 3.

B Study outcomes for the association between caffeine and dementia and/or cognitive function of the included subjects. Pie charts show study outcomes based on population, caffeine consumption dosage and type of caffeine source: positive effect (darker green), positive effect dependent on study characteristics (lighter green), no effect (gray), and negative effect (red [none observed]). Outlined charts indicate a predominant positive outcome.

Caffeine and cognitive function in cognitively impaired individuals

Four studies [22, 35, 40, 41] with a total of 289 subjects assessed the influence of caffeine consumption on cognitive function in cognitively impaired individuals. Cao et al. (2012) [35] assessed concurrent plasma caffeine levels in MCI subjects over a time period of 2–4 years, and observed a reduction in progression to dementia at plasma caffeine levels >1200 ng/ml in this population. Cho et al. (2018) [22] found better global cognitive scores for individuals with PD that consumed coffee, compared to their non-coffee consuming counterparts. Ide et al. (2014) [40] and Ide et al. (2016) [41] both assessed cognitively impaired individuals with AD, VaD, or DLB that consumed green tea powder over a time period of 3 months and 12 months, respectively. Only ‘short-term’ (3 months) green tea consumption was associated with improved cognitive function or reduced progression of cognitive dysfunction.

Taken together, caffeine has a positive effect on cognition in the majority of studies (3/4 (75%) studies including 272/289 (94%) subjects) including cognitively impaired subjects.

Caffeine and study characteristics

Caffeine source

Through categorization of caffeine source that were investigated in each study, we found 29 (48%; 103,321 (67%) subjects) coffee-based studies, 30 (49%; 59,309 (39%) subjects) studies based on tea, 15 (25%; 25,928 (17%) subjects) studies based on multiple caffeinated sources, and 2 (3%; 70 (0.05%) subjects) studies based on pure caffeine (Table 2A–D). Further categorization of tea-based studies revealed 13 (21%; 32,295 (21%) subjects) studies assessing green tea, 7 (11%; 19,635 (13%) subjects) studies assessing black tea and/or oolong tea, and 19 (31%; 37,648 (25%) subjects) studies with other or non-specified tea types (Fig. 3). For the coffee-based studies, we found that 8/29 (28%) studies including 29,515/103,321 (29%) subjects reported a positive association of caffeine consumption on the risk of dementia and/or cognitive decline. Furthermore, 11/29 (38%) studies including 31,681/103,321 (31%) subjects indicated that the outcome was dependent on the quantity of coffee consumed (more positive associations with moderate quantities), sex (more positive for women), age (more positive for older age, 65–74 years), and/or the assessment of short- or long-term association (more protective in the short-term than long-term). The remaining studies on coffee (10/29 (34%); 42,125/103,321 (41%) subjects) reported no association between caffeine and risk of dementia and/or cognitive function. Two studies reported negative associations when long-term effects were assessed [64] or when examining change in habitual consumption [77], but these outcomes shifted toward a positive association when assessing short-term effects and a fixed caffeine consumption frequency and/or concentration over time, respectively.

Table 2A.

Association between coffee-based studies (n = 29) and cognitive decline/dementia

Coffee-based studies
Positive association	No association	Negative association
Al-khateeb et al. 2014 [83]	Arab et al. 2011	Mirza et al. 2014 [64]
	(Sex; men)	(Caffeine consumption duration; long-term)
Arab et al. 2011 [32]	Araújo et al. 2015 [79]	Solfrizzi et al. 2015 [77]
(Sex; women)	(Caffeine consumption quantity and age; ≤1 cup/d or ≥3 cups/d, 35–64 years)	(Change in habitual intake; increased consumption)
Araújo et al. 2015 [79]	Araújo et al. 2016 [63]
(Caffeine consumption quantity and age; 2 –3 cups/d, 65–74 years)	(Caffeine consumption duration; long-term)
Araújo et al. 2016 [63]	Broe et al. 1990 [78]
(Caffeine consumption duration; short-term)
Cho et al. 2018 [22]	Chuang et al. 2019 [66]
	(Caffeine consumption quantity and sex; 2–6 times/wk, men)
Chuang et al. 2019 [66]	Dong et al. 2020 [50]
(Caffeine consumption quantity and sex;≥ 7 times/wk, women)	(Caffeine consumption quantity; < 266.4 g/d)
Dong et al. 2020 [50]	Eskelinen et al. 2009 [28]
(Caffeine consumption quantity; 266.4–495 g/d or ≥495 g/d)	(Caffeine consumption quantity; <3 cups/d and > 5 cups/d)
Eskelinen et al. 2009 [28]	Fischer et al. 2018 [80]
(Caffeine consumption quantity; 3–5 cups/d)Haller et al. 2018 [87]
	Gelber et al. 2011 [24]
(Caffeine consumption quantity; 29–60 cups/months)
Jarvis 1993 [57]	Haller et al. 2018 [87]
	(Caffeine consumption quantity; < 28 cups/months)
Johnson-Kozlow et al. 2002 [38]	Johnson-Kozlow et al. 2002 [38]
(Sex; women)	(Sex; men)
Lee et al. 2017 [67]	Kuriyama et al. 2006 [43]
Lindsay, 2002 [69]	Laitala et al. 2009 [61]
Mirza et al. 2014 [64]	Larsson &Wolk 2018 [86]
(Caffeine consumption quantity and caffeine consumption duration; > 3 cups/d, short-term)
Solfrizzi et al. 2015 [77]	Mirza et al. 2014 [64]
(Caffeine consumption quantity; 1–2 cups/d)	(Caffeine consumption quantity; 1–3 cups/d)
Sugiyama et al. 2016 [46]	Ng et al. 2008 [76]
Valls-Pedret et al. 2012 [23]	Noguchi-Shinohara et al. 2014 [44]
Van Gelder et al. 2007 [62]	Shirai et al. 2020 [45]
(Caffeine consumption quantity; < 4 cups/d)
Wu et al. 2011 [68]	Solfrizzi et al. 2015 [77]
	(Caffeine consumption quantity;< 1 cup/d)
	Tyas et al. 2001 [70]
	Van Gelder et al. 2007 [62]
	(Caffeine consumption quantity;> 4 cups/d)

Bold studies indicate multiple outcomes.

Table 2B.

Association between tea-based studies (n = 30), subdivided into green tea (n = 13), black/oolong tea (n = 7), and other or non-specified tea types (n = 19) and cognitive decline/dementia

Tea-based studies
Green tea
Positive association	No association	Negative association
Ide et al. 2014 [40]	Ide et al. 2016 [41]
Gu et al. 2018	Fischer et al. 2018 [80]
(Caffeine consumption quantity and type of tea source;> 5 times/wk)
Kitamura et al. 2016 [42]	Gu et al. 2018 [51]
	(Caffeine consumption quantity; 1–5 times/wk)
Kuriyama et al. 2006 [43]	Kuriyama et al. 2006 [43]
(Caffeine consumption quantity and type of tea source; ≥ 2 cups/d)	(Caffeine consumption quantity; 4–7 cups/wk)
Lee et al. 2017 [67]	Shen et al. 2015 [53]
	(Type of tea source)
Ng et al. 2008 [76]	Shirai et al. 2020 [45]
	(Caffeine consumption quantity;< once/d)
Noguchi-Shinohara et al. 2014 [44]	Tomata et al. 2016 [47]
(Type of tea source)	(Caffeine consumption quantity;< 2 cups/d)
Shirai et al. 2020 [45]	Xu et al. 2018 [55]
(Caffeine consumption quantity; 2–3 times/d and ≥ 4 times/d)	(Sex and age; women, ≥70 years)
Tomata et al. 2016 [47]
(Caffeine consumption quantity and type of tea source;> 2 cups/d)
Xu et al. 2018 [55]
(Type of tea source, sex and age; men,< 70 years)
Black/Oolong tea
Ng et al. 2008 [76]	Feng et al. 2018 [25]
Shen et al. 2015 [53]	Kuriyama et al. 2006 [43]
(Type of tea source)	(Type of tea source)
	Noguchi-Shinohara et al. 2014 [44]
	(Type of tea source)
	Tomata et al. 2016 [47]
	(Type of tea source)
	Xu et al. 2018 [55]
	(Type of tea source)
Other/non-specified tea type
Arab et al. 2011 [32]	Arab et al. 2011 [32]
(Sex; women)	(Sex; men)
Chen et al. 2012 [49]	Broe et al. 1990 [78]
Chin et al. 2008 [82]	Chuang et al. 2019 [66]
	(Caffeine consumption quantity and sex; 2–6 times/wk, men)
Chuang et al. 2019 [66]	Dai et al. 2006 [36]
(Caffeine consumption quantity and sex; ≥ 7 times/wk, women)
Feng et al. 2012 [75]	Eskelinen et al. 2009 [28]
Huang et al. 2009 [52]	Gu et al. 2018 [51]
(Sex; men)	(Type of tea source)
Jarvis 1993 [57]	Huang et al. 2009 [52]
(Sex; women)
Lee et al. 2017 [67]	Lindsay 2002 [69]
Nurk et al. 2009 [84]	Shen et al. 2015 [53]
	(Caffeine consumption quantity; < 2 cups/d)
Shen et al. 2015	Tyas et al. 2001 [70]
(Caffeine consumption quantity;
2–4 cups/d and ≥4 cups/d)
Wang et al. 2017 [54]	Wang et al. 2017 [54]
(Age;> 60 years)	(Age;> 70 years)
Yang et al. 2016 [56]	Wu et al. 2011 [68]

Bold studies indicate multiple outcomes.

Table 2C.

Association between multiple caffeinated sources (n = 15) and cognitive decline/dementia

Multiple caffeinated sources
Positive association	No association	Negative association
Beydoun et al. 2014	Beydoun et al. 2014 [33]
(Age; ≥70 years)	(Age;< 70 years)
Boot et al. 2013 [34]	Corley et al. 2010 [85]
	(Model; additional adjustments for socioeconomic status and (childhood) IQ)
Cao et al. 2012 [35]	Gelber et al. 2011 [24]
Corley et al. 2010 [85]	Iranpour et al. 2020 [81]
(Model; adjustment for age and sex only)	(Model; multiple additional adjustments)
Driscoll et al. 2016 [37]	Lesk et al. 2009 [58]
Iranpour et al. 2020 [81]	Paganini-Hill et al. 2016 [39]
(Model; no adjustments)	(Caffeine consumption quantity and age; 60–199  mg/d, 70 years)
Maia &de Mendonça 2002 [73]	Ritchie et al. 2007 [71]
(Caffeine consumption quantity and sex;	< 300 mg/d, men)
Paganini-Hill et al. 2016 [39]	Santos et al. 2010 [74]
(Caffeine consumption quantity and age;> 200  mg/d, 90 years)	(Caffeine consumption quantity and sex; < 62  mg/d, men)
Ritchie et al. 2007 [71]	Van Boxtel et al. 2003 [65]
(Caffeine consumption quantity and sex; > 300  mg/d, women)
Santos et al. 2010 [74]	Vercambre et al. 2013 [72]
(Caffeine consumption quantity and sex; > 62  mg/d, women)	(Model; adjustment for age, education and energy from diet only)
Smith 2009 [59]
Vercambre et al. 2013 [72]
(Model; multiple additional adjustments)

Bold studies indicate multiple outcomes.

Table 2D.

Association between pure caffeine (n = 2) and cognitive decline/dementia

Pure caffeine
Positive association	No association	Negative association
Konishi et al.	Walters &Lesk
2018 [48]	2016 [60]

For tea-based studies, we observed 10/30 (33%) studies including 25,381/59,309 (43%) subjects with positive outcomes, 11/30 (37%) studies including 24,556/59,309 (41%) subjects with mixed outcomes dependent on consumed tea source (more positive for green tea), consumed quantity (more positive with moderate quantities), sex (mixed effects), and/or age (mixed effects), Furthermore, 9/30 (30%) studies including 9,372/59,309 (16%) subjects reported no association between tea intake and cognition. No negative associations were found for tea consumption. By classifying the different tea types, we observed proportionally more beneficial associations for green tea (39%) and other/non-specified tea (37%) compared to black/oolong tea (29%). On the other hand, we found that, across most studies (5/7 (71%) studies including 14,603/19,634 (74%) subjects), black/oolong tea was not associated with dementia/ cognitive decline.

Next, we assessed studies that included more than one caffeine source, including coffee, tea, carbonated soft drinks, energy drinks, and foods. Five out of 15 (33%) studies including 6,325/25,928 (24%) subjects reported a protective association and 3/15 (20%) studies including 4,210/25,928 (16%) reported no association between caffeine consumption and cognitive decline. Mixed results were found for 7/15 (47%) studies including 15,393/25,928 (59%) subjects: these studies revealed a dependency of study outcomes according to consumed quantity of caffe-ine, sex, age, and/or covariates in the models. More positive outcomes were found for women compared to men [71, 74], and more positive associations were found for a moderate or higher caffeine quantity (>62mg/d [74], >200 mg/d [39], >300 mg/d [71]). We also found that in studies with mixed caffeine sources, more positive effects were found at ages >70 years, and particularly over 90 years. We found inconclusive findings for the impact of univariate-/basic adjustments or multiple adjustments on cognitive function [72, 81, 85].

Finally, two studies assessed the association of pure caffeine consumption: Konishi et al. (2018) reported better executive function scores, while Walters & Lesk (2016) reported no significant association on cognitive tests.

Our examination of effects in coffee, tea, mixed sources, and pure caffeine-based studies demonstrates that the study outcomes are highly dependent on the caffeine source. Among these caffeine sources, only black/oolong tea seems not to have a protective effect for dementia/cognitive decline. In addition, our data reveal that evidence of a deleterious effect of caffeine consumption on cognitive function is limited.

Caffeine consumption quantity

We assessed the associations between caffeine qu-antity based on the frequency and/or dosage. Of the 61 studies, 48 provided sufficient information to al-low assessment of these associations (Table 3, Fig. 3). Based on pre-specified criteria, the studies were divided into three quantity categories: low caffeine consumption (<100 mg/d) (n = 29, N = 68,470), moderate caffeine consumption (100–400 mg/d) (n = 35, N = 111,776), and high caffeine consumption (>400 mg/d) (n = 14, N = 69,039). For studies with low- and high quantities of caffeine consumption, we mainly found no impact on risk of dementia or cognitive function: positive associations were only observed for 11/29 (38%) and 5/14 (36%) studies respectively. Interestingly, for moderate caffeine consumption, we mainly found beneficial associations with cognitive function (27/35 (77%) studies, that were either dependent (16/35 (46%) studies) or independent of type of caffeine source and/or other study characteristics (11/35 (31%) studies). By further stratifying studies using moderate consumption according to caffeine sources (Table 3), we found that especially consumption of green tea may reduce the risk of dementia and cognitive decline.

Table 3.

Association between caffeine consumption quantity and cognitive decline/dementia

Low caffeine consumption
(<100 mg/d) <1 cup coffee/d or <3 cups tea/d)
Studies	Caffeine source	Quantity	Association (+, /, –)
Arab et al. 2011 [32]	Tea (NS), Coffee	0.57 cups/d	+(women),
		0.95 cups/d	/ (men)
Araújo et al. 2015 [79]	Coffee	≤1 cup/d	/
Araújo et al. 2016 [63]	Coffee	0–1 cup/d	/
Chuang et al. 2019 [66]	Tea (NS), Coffee	2–6 cups/wk	/
Dai et al. 2006 [36]	Tea (NS)	≥3 cups/wk	/
Dong et al. 2020a [50]	Coffee	1–266.4 mg/d	/
Feng et al. 2018 [25]	Tea (Black)	1 cup/wk	/
Gu et al. 2018 [51]	Tea (Green), Tea (NS)	1–5 times/wk	/
Haller et al. 2018 [87]	Coffee	<28 cups/month	/
Ide et al. 2014 [40]	Tea (Green tea powder)	2 g/d (<100 mg/d caffeine)	+
Ide et al. 2016 [41]	Tea (Green tea powder)	2 g/d (<100 mg/d caffeine)	/
Iranpour et al. 2020 [81]	Multiple sources	11–102 mg/d	/
Kitamura et al. 2016 [42]	Tea (Green)	1–6 cups/wk	+(univariate model),
			/ (multiple additional adjustments)
Kuriyama et al. 2006 [43]	Tea (Green), Tea	<1 cup/d	/
	(Black/oolong), Coffee
Lee et al. 2017a [67]	Tea (Green), Tea	>3 cups/wk	+
	(Black/oolong), Coffee
Lesk et al. 2009 [58]	Multiple sources	Mean: 70.3 (±36.2) mg/d	/
Maia &de Mendonça, 2002 [73]	Multiple sources	Mean: 73.9 (±97.9) mg/d	+
Ng et al. 2008 [76]	Tea (Green), Tea	<1 cup/d	+
	(Black/oolong), Coffee		/
Noguchi-Shinohara et al. 2014 [44]	Tea (Green)	<1 cup/d	+
	Tea (Black), Coffee		/
Paganini-Hill et al. 2016a [39]	Multiple sources	50–199 mg/d	/
Ritchie et al. 2007a [71]	Multiple sources	100–200 mg/d	/
Santos et al. 2010 [74]	Multiple sources	22–62 mg/day	/
Shen et al. 2015 [53]	Tea (Black), Tea (Green)	<2 cups/d	/
Shirai et al. 2020 [45]	Tea (Green)	2–3 times/d	+
Solfrizzi et al. 2015a [77]	Coffee	1 cup/d	/
Tomata et al. 2016 [47]	Tea (Green),	1–2 cup/d	/
	Tea (Black/oolong)
Valls-Pedret et al. 2012 [23]	Coffee	Median: 21 ml/d	+
Wu et al. 2011 [68]	Coffee	>1 cup/wk	+
	Tea (NS)		/
Xu et al. 2018 [55]	Tea (Green)	>3 cup/wk	+(men, particularly < 70
	Tea (Black/oolong)	years),/ (women)
Moderate caffeine consumption
(100–400 mg/d)
1–4 cups coffee/d or 3–10 cups tea/d
Araújo et al. 2015 [79]	Coffee	2–3 cups/d	+(65–74 years),
			/ (35–64 years)
Araújo et al. 2016 [63]	Coffee	1–3 cups/d	/
Beydoun et al. 2014 [33]	Multiple sources	Mean: 132 mg/d	+(≥70 years),
			/ (< 70 years)
Broe et al. 1990 [78]	Tea (NS)	>4 cups/d	/
Chin et al. 2008 [82]	Tea (NS)	Mean: 4.46 cups/d	+
Chuang et al. 2019a [66]	Tea (NS), Coffee	≥7 cups/wk	+(all subjects and women),
			/ (men)
Corley et al. 2010 [85]	Multiple sources	Mean: 182.5 mg/d	+(adjustment for age and),
			sex/ (multiple additional)
			adjustments
Dong et al. 2020a [50]	Coffee	266.4–295 mg/d	+
Driscoll et al. 2016 [37]	Multiple sources	Mean: 261 mg/d	+
Eskelinen et al. 2009a [28]	Coffee	3–5 cups/d	+
Feng et al. 2018a [25]	Tea (Black)	>1 cup/d	/
Gelber et al. 2011 [24]	Coffee, Multiple sources	115.5–188.0 mg/d	/
Gu et al. 2018 [51]	Tea (Green)	>5 times/wk	+
	Tea (NS)		/
Haller et al. 2018 [87]	Coffee	29–60 cups/mo	+
Iranpour et al. 2020 [81]	Multiple sources	>209 mg/d	+(univariate model),
			/ (multiple additional adjustments)
Johnson-Kozlow et al. 2002 [38]	Coffee	Mean: 3 cups/d	+(women),
			/ (men)
Kitamura et al. 2016a [42]	Tea (Green)	>1 cup/d	+
Konishi et al. 2018 [48]	Pure caffeine	200 mg/d	+
Kuriyama et al. 2006 [43]	Tea (Green)	≥2 cups/d	+
	Tea (Black/oolong),		/
	Coffee
Larsson &Wolk, 2018a [86]	Coffee	1.0–4.9 cups/d	/
Lindsay, 2002 [69]	Coffee	>1 cup/d	+
	Tea (NS)		/
Mirza et al. 2014 [64]	Coffee	1–3 cup/d	/
Ng et al. 2008 [76]	Tea (Green),	>1 cup/d	+
	Tea (Black/oolong)		/
	Coffee
Noguchi-Shinohara et al. 2014 [44]	Tea (Green)	>1 cup/d	+
	Coffee		/
Paganini-Hill et al. 2016 [39]	Multiple sources	>200 mg/d	+(> 90 years),/(> 70 years)
Ritchie et al. 2007 [71]	Multiple sources	200–300 mg/d	/
Santos et al. 2010a [74]	Multiple sources	>62 mg/day	+(women),
			/ (men)
Shen et al. 2015 [53]	Tea (Black)	≥4 cups/d	+
	Tea (Green)		/
Shirai et al. 2020 [45]	Tea (Green)	≥4 times/d	+
	Coffee	≥2 times/d	/
Smith, 2009 [59]	Multiple sources	Mean: 140 mg/d	+
Solfrizzi et al. 2015 [77]	Coffee	1–2 cups/d	+
Sugiyama et al. 2016 [46]	Coffee	1–2 cups/d	+
Tomata et al. 2016 [47]	Tea (Green)	≥5 cups/d	+
	Tea (Black/oolong)		/
Van Gelder et al. 2007 [62]	Coffee	1–4 cups/d	+
Walters &Lesk, 2016 [60]	Pure caffeine	200 mg/d	/
High caffeine consumption
(>400 mg/d)
>4 cups coffee/d,>10 cups tea/d
Araújo et al. 2015a [79]	Coffee	≥3 cups/d	/
Araújo et al. 2016a [63]	Coffee	≥3 cups/d	+(short-term),/ (long-term)
Broe et al. 1990 [78]	Coffee	≥4 cups/d	/
Dong et al. 2020 [50]	Coffee	≥495 mg/d	+
Eskelinen et al. 2009 [28]	Coffee	>5 cups/d	/
Gelber et al. 2011 [24]	Coffee	415–2673 mg/d	/
	Multiple sources
Haller et al. 2018a [87]	Coffee	61–168 cups/mo	/
Laitala et al. 2009 [61]	Coffee	Mean: 5.3 cups/d	/
Larsson &Wolk, 2018a [86]	Coffee	≥5.0 cups/d	/
Mirza et al. 2014a [64]	Coffee	>3 cups/d	+(short-term),
			/ (long-term)
Ritchie et al. 2007a [71]	Multiple sources	>300 mg/d	+(women),
			/ (men)
Van Boxtel et al. 2003 [65]	Multiple sources	Median: 5–6 cups/d	/
Van Gelder et al. 2007 [62]	Coffee	>4 cups/d	/
Vercambre et al. 2013 [72]	Multiple sources	>371 mg/d	+

^aCategorization in this group due to different categories used in the study.

Confounding factors

Most studies adjusted for age and sex, and in a subset of studies additional model adjustments were made for factors like hypertension, diabetes mellitus, hyperlipidemia, education, APOE genotype, smoking, alcohol, physical activities, body mass index (BMI), socioeconomic status, and global cognition (MMSE). Some studies reported an impact of confounding factors on outcomes.

For seven studies [32, 38, 52, 55, 66, 71, 74], outcomes were dependent on sex. These studies reported that beneficial associations are predominantly found in women (5/7 studies). In line with these findings, two studies with only female participants [37, 72] reported positive associations and two out of three studies with only male participants [24, 25, 62] reported no associations.

Four studies indicated that positive associations are dependent on age. These studies reported positive associations between caffeine consumption and dementia and/or cognitive function at older ages (65–74 years versus 35–64 years [79], >70 years versus <70 years [33], 90 years versus 70 years [39]). However, two other studies indicated the reverse, an effect at younger age (>60 years versus >70 years [54]) or that effects were particularly found at ages <70 years old [55].

Furthermore, Mirza et al. (2014) [64] and Araújo et al. (2016) [63] found different outcomes depending on the time of follow-up. Short-term follow-up (within 4 years) revealed positive associations, while the association was negative at long-term follow-up (>4 years) [64] and absent in another study implementing a long-term follow-up (5.5 years) [63].

Corley et al. (2010) [85] observed protective associations between caffeine and cognitive function when adjusting for age and sex, but when additional adjustments were made for socioeconomic status or social class and (childhood) IQ, the association did not reach the threshold for statistical significance. Similar results were observed by Iranpour et al. (2020) [81], who reported a positive association in a univariate model but no association in models where adjustments for factors like sex, age, race/ethnicity, education, and marital status, or self-rated health, disease history, and depression were made. Vercambre et al. (2013) [72], on the other hand, only found a positive association when adjusting for alcohol consumption, physical activity, BMI, and smoking, but not when only adjusting for age, education, and diet. Moreover, this study found a more pronounced positive association with caffeine when it was supplemented with vitamin B.

DISCUSSION

In this systematic review, we assessed the association between caffeine and 1) the risk of dementia and/or cognitive decline and 2) cognitive function in individuals with impaired cognition (i.e., MCI or dementia). The number of studies showing positive associations (dependent or independent of study cha-racteristics) was 46/57 (81%) including 111,926/153,070 (73%) subjects, indicating that caffeine has a beneficial effect on the risk of dementia/cognitive decline. We also found more positive results (3/4 (75%) studies including 272/289 (94%) subjects) for studies that included subjects with MCI, or any type of dementia, indicating that caffeine also has a beneficial effect in cognitively impaired subjects. Furthermore, we observed that various study characteristics affect the reported associations of caffeine such that moderate consumption seems to be more beneficial than low- or high quantities, and coffee, green- and other/non-specified tea, and multiple caffeinated sources are more beneficial than other caffeine sources like black/oolong tea. Effects were also found to be more pronounced in women compared to men, and many studies reported mixed outcomes based on other factors like age and follow-up time. Across all studies, we observed only two studies with a negative effect, suggesting that caffeine is unlikely to negatively affect cognition or dementia risk. This review highlights that dietary factors may influence risk of cognitive decline and dementia, and may also aid the future development of caffeine-based intervention studies, which might serve as a cost-effective alternative or add-on to other non-pharmacological or pharmacological treatments against cognitive decline and dementia (e.g., physical activity [88]).

Potential mechanisms

Results from this review suggest that caffeine effects are dependent on the caffeine source and quantity. Several explanations exist for this outcome. First, different types of caffeine sources contain different levels of caffeine [29], and low dosages might be inadequate to convey positive effects while with excessive dosages the negative effects (e.g., anxiety) might outweigh the positive effects. There might also be individual variability in the physiological response to caffeine (e.g., due to genetic factors that influence responsiveness of A_2A receptors), which would result in differential effects of the same dose of caffeine across individuals [14, 89, 90]. Furthermore, physiological effects of other substances than caffeine that are contained within the caffeine source (e.g., coffee) may influence or strengthen the caff-eine response, by affecting the kinetics of caffeine in the body and the response of adenosine receptors, or have a caffeine independent effect that influen-ces cognitive performance [91]. For example, various sources of caffeine contain antioxidants, which have been found to play a role in protecting against oxi-dative stress, and may thereby help in preventing cognitive deterioration [92]. Coffee displays antioxidative effects through chlorogenic acid and poly-phenols [93]. Tea displays antioxidative effects through tea catechins and theaflavins, and green tea exhibits higher antioxidative effects than black or oolong tea [94]. Varying antioxidative mechanisms or degrees of antioxidative effects might contribute to the differences in study effects according to caf-feine sources observed in this report (i.e., more effects in green compared to black/oolong tea). However, further research is needed on the effect of antioxidants as studies have also reported no effect of antioxidants on cognitive function, but rather on mood [95]. Caffeine may also lead to better cognitive function and memory indirectly through an increase in alertness and wakefulness [12], and by influencing sleep and impulsivity [14, 96].

Caffeine has also been found to influence neural and vascular activity such as vasoconstriction and reduction in cerebral blood flow (CBF). Reduction in CBF leads to an increased oxygen extraction from the blood to cerebral structures in the brain [97], which, in turn, enhances cognitive performance. It seems possible that a sufficient quantity must be ingested in order to produce this effect. On the other hand, excessive caffeine consumption could lead to (acute) caffeine overdose, which could convey negative effects such as reactive oxygen species formation [98], that outweigh the positive, or indirect negative symptoms that could influence cognitive function such as restlessness, anxiety, agitation, insomnia, and headache [16].

This review revealed incongruent outcomes for other confounding factors such as sex, age, and follow-up time. It seems that caffeine consumption is particularly beneficial for cognitive function in women in comparison with men. In general, inconsi-stent results for women and men might be explained through sex-based biological variations such as tes-tosterone and estrogen hormone levels [99]. Furthermore, four studies reported an outcome that was dependent on age, but it remains to be determined at what age caffeine has the most beneficial effect as some studies reported greater effects in older subjects, while others reported greater effects in younger subjects. Follow-up time was also found to influence outcomes in two studies. These studies both reported beneficial associations at a short follow-up time, while no effects were observed at a long-term follow up. This suggests that the beneficial effects of caffeine might be temporary.

Strengths and limitations

The main strength is that we performed a systematic review and assessed all available studies, regardless of study design. Thereby, we were able to include more studies than have previously been in-cluded in other reviews and meta-analyses [100–103]. However, there are also limitations that need to be considered when interpreting this review. First of all, it is important to highlight that, in the second-ary analysis on cognitively impaired individuals, we were able to assess only four studies, and that these studies included individuals with different types of cognitive impairments, various caffeine sources and different study designs. Also, one out of four studies included patients with PD, for which the degrading underlying mechanisms are different compared to patients with dementia or MCI. Secondly, our app-roach of providing this systematic review did not allow us to perform formal statistical analyses to assess the effects of caffeine quantitatively, or stati-stically assess modifying effects. This lack of quantitative assessments means our findings were based exclusively on overall study outcomes. Thirdly, our interpretation of the included studies relied on data provided in the paper, and we did not contact the authors to provide additional information because of the wide inclusion timeframe of this review (1990–2020). As a result, not all studies could be included when assessing study characteristics. For example, accurate information on caffeine quantity was not al-ways provided. Furthermore, many studies employed self-reported caffeine consumption data resulting in a high risk of bias due to deviations from the intended intervention. Finally, information on reporting of funding sources and conflicts of interests were not considered as possible confounders in the analyses.

Conclusion

Our findings indicate that caffeine beneficially affects cognitive function and risk of dementia and that this effect is dependent on the type of caffeine source (e.g., more effects for coffee and green tea), quantity (more effects with moderate quantities), and sex (more effects in female subjects). Furthermore, we found that other factors such as age and follow-up time might influence effects and it is important for future studies to examine, and account for, these confounders. Ideally, future investigation should implement a randomized-controlled trial design, which would allow for quantitative assessme-nts of effects across studies. Future studies including various dosage levels could additionally help to extend our findings regarding the most beneficial caffeine dosage by accurately determining the optimal caffeine quantity to effect cognitive decline and risk of dementia. Furthermore, it would be interesting to map genetic factors that influence response to caffeine (e.g., A_2A receptor haplotype) in future studies, as differences in responsiveness to caffeine could influence effects of caffeine on cognition. These in-sights may help in tailoring cost-effective lifestyle interventions, and possibly even aid in the development of pharmacological interventions that combat cognitive decline and dementia.

DISCLOSURE STATEMENT

Authors’ disclosures available online (https://www.j-alz.com/manuscript-disclosures/20-1069).