Coffee is the leading beverage after water worldwide, and its trade exceeds US$
10 billion [1
]. Caffeine is present in many dietary sources consumed around the world, such as in coffee, tea, candy bars and cocoa beverages. The amount of caffeine ranges quite widely between these various foods, with coffee representing a major source of intake (71–220 mg caffeine/150 ml) [2
]. Coffee has been shown to exert beneficial effects toward human health, including cardiovascular health, several types of cancer and neurodegenerative diseases [5
], due to prevailing mechanisms such as inhibition of oxidative stress, regulation of DNA repair, phase II enzymatic activity, apoptosis and inflammation [6
]. However, epidemiological evidence has shown that pregnant women and their offspring might be subjected to detrimental effects of caffeinated coffee [5
As life expectancy increases, age-related cognitive decline can be a major health challenge for the elderly population [7
], cognitive health has become an important public health issue for America’s aging population [8
]. The process from cognitive decline to dementia is continuous and irreversible, and there is no effective treatment for dementia so far, the therapeutic value of drugs currently used is limited. Thus, developing measures to reduce risk for low cognitive performance as well as treatments of diagnosed dementia occupy a high priority in society.
A number of epidemiological studies have demonstrated an association between higher coffee consumption and better cognitive performance [9
]. Cognitive benefits from coffee consumption were typically attributed to caffeine. Caffeine is an antagonist of A1 and A2A adenosine receptors in the central nervous system and is known to have positive effects on attention, arousal, mood and vigilance [14
]. Epidemiological studies have reported that caffeine was associated with cognitive impairment [17
], suggesting that caffeine had a protective effect on cognitive performance. Furthermore, some studies of the association between caffeine and cognitive performance have also noted differential associations by gender, and the results were inconsistent [19
]. However, few studies have explored the associations of decaffeinated coffee and caffeine intake from coffee with cognitive performance, and the results were inconsistent. Some studies have indicated that decaffeinated coffee might bring about some improvements to cognitive performance [23
], while other studies have found that there was no significant association between decaffeinated coffee and cognitive function [25
]. Moreover, some studies have suggested that caffeine from coffee was associated with cognitive performance [19
], while other studies have reported null associations [28
Therefore, we analyzed a nationally representative sample of older adults aged 60 years or older from the National Health and Nutrition Examination Survey (NHANES) to investigate the associations of total coffee, decaffeinated coffee, caffeinated coffee and caffeine intake from coffee with cognitive performance.
Of all the participants, there were significant differences (p
< 0.01) between individuals with low cognitive performance and normal cognitive performance in the distribution of race, educational level, poverty-income ratio, smoking status, diabetes, stroke, total energy intake, coffee intake, caffeinated coffee intake and caffeine intake from coffee among the CERAD test, Animal Fluency test and DSST (Table 1
). As can be seen in the tables, those who reported low cognitive performance were more likely to be non-black, current smokers, have lower educational level, poverty-income ratio, less coffee intake, less caffeinated coffee intake, less caffeine intake and higher prevalence of diabetes and stroke than those who reported normal cognitive performance. Participants in the low cognitive performance group who took the CERAD and Animal Fluency test were more likely to be older. For the DSST and Animal Fluency test, the prevalence of hypertension in people with low cognitive performance was significantly higher than that of people with normal cognitive performance, and the alcohol drinking rate was lower in people with low cognitive performance than people with normal cognitive performance. People in the low cognitive performance group with the CERAD and DSST tests tended to be male, whereas the normal cognitive performance people were more likely to be female.
shows the associations between total coffee intake and different dimensions of cognitive performance. Compared to those reporting no coffee consumption, those who reported 266.4–495 (g/day) had a crude odd ratio (OR) with 95% confidence interval (CI) of 0.74(0.50–0.91) for DSST score. After adjustment for age and gender, total coffee intake was still associated with cognitive performance. In Model 3, compared to those reporting no coffee consumption, those who reported 266.4–495 (g/day) had a multivariate-adjusted OR (95% CI) of 0.56(0.35–0.89).
presents the associations of caffeinated coffee and decaffeinated coffee with cognitive performance. Compared to those reporting no caffeinated coffee consumption, those who reported ≥384.8 (g/day) had a crude OR with 95% CI of 0.58(0.42–0.81) for DSST score. After adjustment for age and gender, caffeinated coffee intake was still associated with cognitive performance. In Model 3, compared to those reporting no caffeinated coffee consumption, those who reported ≥384.8 (g/day) had a multivariate-adjusted OR (95% CI) of 0.68(0.48–0.97). No significant association was observed between decaffeinated coffee and different dimensions of cognitive performance. In sensitivity analysis, the association of caffeinated coffee with DSST score was still significant, and the association of decaffeinated coffee with cognitive performance was not significant (Supplementary Table S2
shows the associations between caffeine intake from coffee and different dimensions of cognitive performance. Compared to the lowest quartile of caffeine intake from coffee, the crude OR (95% CI) of the quartile (Q) three was 0.60(0.40–0.91) for CERAD test score. After adjustment for age and gender, caffeine intake was still associated with cognitive performance. In Model 3, compared to the lowest quartile of caffeine intake from coffee, the multivariate adjusted OR (95% CI) of the quartile (Q) three was 0.62(0.38–0.98). In sensitivity analysis, in the fully adjusted model, the negative associations of caffeine from coffee with CERAD test score and DSST score were significant (Supplementary Table S3
We also evaluated the associations between caffeine intake from coffee and cognitive performance among men and women, separately, to assess potential differences by gender (Supplementary Table S4
). For women, caffeine intake was associated with CERAD test score and DSST score, the corresponding ORs (95% CIs) were 0.34(0.17–0.65) and 0.39(0.20–0.76) in Model 3. For men, there was no significant association between caffeine intake and different dimensions of cognitive performance in Model 3. A statistically significant interaction was noted between caffeine from coffee and gender in the CERAD test in the model that adjusted for the same covariates (p
The results of linear regression analyses of associations between total coffee consumption, caffeinated coffee consumption, decaffeinated coffee consumption, caffeine from coffee and cognitive performance is shown in the Supplementary Materials (Tables S5–S7)
. In the fully adjusted model, there was significant association between coffee consumption and DSST score (β = 0.0017, 95% CI: 0.0001–0.003). Moreover, the associations of caffeinated coffee with Animal Fluency test score (β = 0.0006, 95% CI: 0.00001–0.0013) and DSST score (β = 0.0021, 95% CI: 0.0003–0.004) were significant in Model 3. No significant association was observed between decaffeinated coffee and different dimensions of cognitive performance. Furthermore, there was a significant association between caffeine from coffee and CERAD test score (β = 0.0025, 95% CI: 0.0001–0.0049) in Model 3.
, Figure 3
and Figure 4
depict the results of the restricted cubic spline analyses. We found a suggestion of L-shaped associations of total coffee intake and caffeinated coffee intake with DSST score. The prevalence of low cognitive performance decreased with increasing intake of total coffee and caffeinated coffee and showed a nonlinear dose–response relationship (p total coffee
for nonlinearity = 0.039, p caffeinated coffee
for nonlinearity = 0.023). We also found a suggestion of L-shaped associations between caffeine intake from coffee and CERAD test score. The prevalence of low cognitive performance decreased with increasing intake of caffeine and showed a nonlinear dose–response relationship (p caffeine
for nonlinearity = 0.032).
In this study, we combined data from NHANES 2011–2012 and 2013–2014 and included 2513 Americans aged 60 years or older. In the fully adjusted model, the associations of total coffee, caffeinated coffee and caffeine intake from coffee with DSST score and CERAD test score were significant, and L-shaped dose–response relationships were also detected. No significant association was observed between decaffeinated coffee and different dimensions of cognitive performance. In sensitivity analyses, the associations of caffeinated coffee and caffeine from coffee with DSST score and CERAD test score were still significant by excluding decaffeinated coffee consumers. The association of decaffeinated coffee with cognitive performance was not significant by excluding caffeinated coffee consumers. In stratified analyses, higher levels of caffeine intake from coffee were associated with higher CERAD test score and DSST score in women but not in men.
Our finding about coffee consumption was partially consistent with the findings from some previous studies [11
]. A population-based study of 145 community-based older individuals [11
] found a positive effect of coffee on cognitive performance. In addition, a recent 30-year follow-up study of 8000 Japanese-American men [12
] suggested that coffee intake might protect against Parkinson’s disease. Alzheimer’s and Parkinson’s diseases are both neurodegenerative, approximately thirty percent of Parkinson’s disease patients might develop an Alzheimer’s-like dementia and thirty percent of Alzheimer’s patients might develop Parkinson’s-like changes [52
]. In a large, population-based study of 9003 British people, Jarvis [53
] found a significant positive trend between coffee intake and cognitive performance. Furthermore, studies of the association between coffee and cognitive performance also indicated that although reduced risk was related to coffee consumption in men [51
], the effect was more pronounced in women [19
], whereas some studies [54
] showed null or adverse associations. A population-based Rotterdam study [55
] of 2914 participants in a five-year follow-up, and a cohort study of 14,563 participants (35–74 years old) conducted by Araújo [56
], showed null or adverse effects of coffee consumption on cognitive performance.
We also found that caffeinated coffee and caffeine from coffee were associated with cognitive performance, which were consistent with previous studies. A population-based cohort study of 7017 community-based older individuals [19
] showed that caffeine intake from coffee appeared to reduce cognitive decline. In a placebo-controlled cross-over design [27
], caffeine intake from coffee was also found to have a protective effect on cognitive performance. Moreover, a meta-analysis of eleven observational studies [58
] also suggested a positive effect of caffeine from coffee on cognitive performance, with a summary relative risk (RR) of 0.84 (95% CI: 0.72–0.99, I2 = 42.6%). However, a meta-analysis of observational studies found that caffeine intake from coffee was not associated with the risk of cognitive disorders [28
In addition, no significant association between decaffeinated coffee and cognitive performance was found in our study, likely reflecting lower statistical power for these analyses or due to the small number of participants. The finding was partially consistent with those of the population-based study of 1528 elderly people, conducted by Johnson-Kozlow [25
], which found a positive effect of caffeinated coffee on cognitive performance, and there was no significant association between decaffeinated coffee intake and cognitive function. Moreover, a placebo-controlled trial of sixty older individuals suggested that no significant association was observed between decaffeinated coffee and cognitive function [26
]. In contrast, a randomized placebo-controlled study indicated that decaffeinated coffee might have a protective effect on cognitive performance [23
]. An animal trial conducted by Jang et al. also provided evidence that decaffeinated coffee might prevent memory impairment in humans [24
The mechanisms of the relationship between caffeine intake and cognitive performance remained unclear, but there have been several possibilities. Caffeine may have the ability to induce mRNA and protein expression and mediate NF-E2-related factor 2-Antioxidant Response Element (Nrf2-ARE) pathway stimulation, which could improve the overall antioxidant capacity of the body and thus contribute to ameliorating oxidative stress, inflammation and carcinogenesis [1
]. Moreover, Riedel et al. [27
] reversed the effects of scopolamine through the administration of 250 mg of caffeine and concluded that caffeine acted through cholinergic pathways and specifically enhanced memory. Furthermore, caffeine and its metabolites helped in proper cognitive performance. Coffee lipid fraction containing cafestol and kahweol played a protective role against some malignant cells by regulating the detoxifying enzymes.
The differences found between men and women indicated that women were more vulnerable to the effects of caffeine than men. The elimination half-life of caffeine ranged from three to seven hours; however, elimination was about twenty percent shorter because of more rapid biotransformation among women [59
]. Research by Carrillo [60
] indicated that women were more likely than men to experience acute toxic reactions, such as restlessness, palpitation, muscle tremor and dizziness, after taking high doses of caffeine. Therefore, gender differences may be due to pharmacodynamic differences in sensitivity of men and women to caffeine effects. In another study, Relling et al. [61
] indicated that healthy women had higher levels of xanthine oxidase activity than did men after ingesting equal amounts of caffeine, suggesting that men and women metabolized caffeine differently. A randomized controlled trial also provided evidence of different responses of men and women to caffeine, which may be mediated by changes in circulating steroid hormones [62
Our study presents several advantages. A major strength was the use of a large nationally representative sample of older adults in the United States. In terms of survey methods and quality control, the NHANES was high quality. In addition, wide ranges of potential confounders were controlled to provide a better estimate of the association of coffee and caffeine intake with cognitive performance. Moreover, we investigated the dose–response relationship of coffee and caffeine consumption with cognitive performance.
We acknowledge several limitations of our study. Primarily, as a cross-sectional study, these associations cannot necessarily be considered as causality, so it was difficult to generalize the results of this study to the causal relationship from coffee consumption to cognitive performance. Furthermore, the cognitive tests, chosen for ease of administration, availability and use in other surveys, did not cover all domains of cognitive function. Adults who performed well in one domain may not perform well in another domain. What is more, we cannot rule out the co-linearity effect in this study. Finally, the dietary data was obtained from two 24-hour dietary recall interviews, which did not accurately reflect individuals’ usual intake, but some studies have shown that two 24-h recalls might be sufficient to assess the daily dietary intake [63