Low Intake of Total Antioxidant Nutrients as a Risk Factor for Incident Dementia in Older Adults: The Shanghai Aging Study

Dementia is a neurological disorder with the symptoms of a significant decline from one’s previous level of cognition that causes interference in domestic, occupational, or social functioning [1]. Advancing age, genetic profile, and systemic vascular disease are major risk factors for developing dementia [2]. Therefore, impaired cognition has become a major challenge to public health as the population ages [3]. Globally, about 50 million people live with dementia, and the number of cases is increasing to 152 million by 2050. According to a national epidemiological study, about 15 million people aged 60 years or older in China suffered with dementia with a prevalence rate of 6.0%, which is similar to that in most parts of the world (5.5–7.0%), but is higher than that in sub-Saharan Africa (5.5%) and central Europe (5.2%) and lower than that reported in Latin America (8.4%) and southeast Asia (7.6%) [4]. The pathophysiology of dementia include defective beta-amyloid protein metabolism, abnormalities of adrenergic, glutamatergic, dopaminergic, and serotonergic neurotransmission, and the involvement of inflammatory, hormonal, and oxidative pathways [5].

The common antioxidant nutrients (ANs) include carotene, vitamin E, vitamin C, flavonoids, lutein, and some microelements like selenium. These ANs usually come from tomato, cucumber, cabbage, laver, lettuce, and other leafy green foods. It was reported that higher dietary AN intake might prevent neurodegenerative diseases (e.g., AD) by mitigating the destruction of free radicals on neurons [6, 7]. Large epidemiologic studies have shown that a greater adherence to the Mediterranean diet is associated with elevated non-enzymatic antioxidant capacity levels [8, 9]. Different studies have shown the ability of diet to modulate plasma non-enzymatic antioxidant capacity following acute consumption of plant foods in humans [10‒13]. Previous literature reported the impact of AN on cognitive impairment in the older population, but the conclusions were inconsistent. Some studies indicated that AN intake (such as vitamin C, vitamin E, and polyphenols, in particular flavonoids) reduced the risk of cognitive decline and dementia [14‒16] while others did not find significant associations [17‒19]. Few studies just focused on the intake of individual AN rather than the combined effects of total AN. On the other hand, most previous studies were from Western countries, and evidence from the Asian population, especially the Chinese population, is lacking.

As a community-based cohort study, the Shanghai Aging Studyʼs design, diagnostic criteria, and operational procedures are similar to most cohort studies focusing on cognitive function in high-income countries [20]. The current study intended to verify the hypothesis that dietary intake of total AN was associated with incident dementia, by analyzing the 5-year follow-up data of the Shanghai Aging Study.

From 2010 to 2011, the Shanghai Aging Study, an observational prospective cohort study, recruited participants aged 60 years or older from a community in urban Shanghai, China. Inclusion criteria of the study have been published elsewhere [20]. In the current study, we further included participants who were (1) dementia free at the baseline, (2) able to provide the baseline data of their daily dietary intake, and (3) able to complete the follow-up interview.

This study was approved by the Medical Ethics Committee of Huashan Hospital, Fudan University, Shanghai, China (No. 2009-195). Written informed consent was obtained from all the participants at the baseline recruitment.

At baseline, a 111-item food frequency questionnaire (FFQ) (with 81 groups of foods in 8 categories), which recorded the frequency and amount of dietary intake over the past 1 year for each participant, was used to measure the daily dietary intake. The FFQ was adjusted to suit the local diet in Shanghai, such as salted foods, fermented foods, soy foods, allium vegetables, and leafy vegetables, and has been validated in the Chinese population [21, 22]. Average daily intakes of dietary AN, i.e., carotene (total-carotene in the current study), vitamin C, vitamin E, lutein, and flavonoids, and energy, were calculated based on the China Food Composition, 2nd Edition [23], and literature of Chinese food composition [24‒26]. Carotene mainly comes from dark green, red, or yellow fruits and vegetables, such as carrots, broccoli, and spinach. Vitamin C mainly comes from fruits, vegetables, and other plant foods, such as apples, oranges, balsam pear, and celery. Vitamin E mainly comes from various oils, cereals, nuts, meat, milk, eggs, and other foods. Lutein has a wide range of food sources, which often coexist with zeaxanthin in nature. It is mainly from orange vegetables and fruits, such as corn, pumpkin, and purple cabbage. Flavonoids are widely found in natural plants, such as tomato, carrot, strawberry, and tea. In the current study, total AN intake was defined as the sum of carotene, vitamin C, vitamin E, lutein, and flavonoids intake.

At baseline, each participant was administered the neurological examinations and Lawton and Brody Activity of Daily Living (ADL) scale [27] by neurologists and research nurses at Huashan Hospital. The person was considered to have normal functioning if their ADL score was greater than 16 [27]. Study psychometrists conducted a battery of neuropsychological tests for each participant. Neuropsychological tests included (1) Mini-Mental State Examination [28]; (2) Conflicting Instructions Task (Go/No Go Task) [29]; (3) Stick Test [30]; (4) Modified Common Objects Sorting Test [30]; (5) Auditory Verbal Learning Test [31]; (6) Modified Fuld Object Memory Evaluation [32]; (7) Trail Making Tests A and B [33]; (8) RMB (Chinese currency) test [34], covering the global cognition and domains of memory, language, attention, executive function, and spatial construction. All the tests were administered according to the education years of each participant: tests (1) to (5) and (7) for participants with ≥6 years of education, while tests (1) to (4), (6) and (8) for those with <6 years of education [35]. A detailed description and normative data of these tests were reported previously [35, 36]. Diagnosis of dementia was made based on the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV), by reviewing the neuropsychological assessments, ADL, and neuropsychological tests at the baseline and the follow-up interviews.

Based on the knowledge in nutrition and neurology, we defined confounders related to cognition and diet, i.e., age, gender, education year, obesity, apolipoprotein E (APOE)-ε4 allele, medical history of hypertension, diabetes, and depression. Body mass index (BMI) was calculated by the body weight in kilograms divided by the squared height in meters. BMI ≥27.5 kg/m² was defined as obesity by the World Health Organization criteria [37]. Comorbidities of type 2 diabetes and hypertension were self-reported and confirmed by the medical records of each participant. Depressive symptoms within the past 1 week were defined if the Center for Epidemiologic Studies Depression Scale (CES-D) ≥16 [38].

DNA was extracted from blood or saliva samples of participants. TaqMan SNP method was used for the APOE genotyping [39]. APOE-ε4 positive was defined as the presence of at least one ε4 allele.

Participants without dementia at baseline were invited for the follow-up interview between April 1, 2014, and December 31, 2016. The same neurological and neuropsychological evaluation process that was used at the beginning of the study was repeated during the follow-up interview. Any new cases of dementia detected during the follow-up period were diagnosed using the same methods and standards as those used initially.

Continuous variables that conform to normal distribution were presented as the mean and standard deviation, and the others were presented as the median and interquartile range. Categorical variables were presented as numbers and frequencies (%). The Student’s t test, Wilcoxon rank-sum test, Pearson χ² test, and Fisher’s exact test were used to compare continuous and categorical variables where they were appropriate.

The incidence rate of dementia was calculated as the number of newly onset dementia cases divided by the total person-years of follow-up. The Kaplan-Meier curve was plotted for the cumulative incidence rate of dementia by follow-up time.

We used the Cox proportional hazards regression model and the restrictive cubic spline to determine the proper cut-off values to categorize the average daily intake of total AN with the adjustment of age, gender, education year, obesity, APOE-ε4, medical history of hypertension, diabetes, depression, baseline MMSE, and total energy intake. As shown in Figure 1, the gray shadow represents the 95% confidence interval (CI) of hazard ratio (HR) in total AN intake, and the black solid line represents the estimated value of HR. The intersection of the short-dashed lines indicated that, when the total AN intake was less than 112 mg/day, the lower bound of the 95% CI of HR was higher than 1, suggesting a significant increased risk for dementia.

We defined the highest quartile of intake groups as the reference groups and estimated the HRs of incident dementia in the lowest quartile of intake groups of carotene, vitamin C, vitamin E, lutein, and flavonoid, respectively, by the multiple (5 times) independent multivariate Cox regression model adjusted for age, gender, education year, obesity, APOE-ε4, medical history of hypertension, diabetes, depression, baseline MMSE, and total energy intake. A multivariate Cox regression model adjusting for age, gender, education year, obesity, APOE-ε4, medical history of hypertension, diabetes, depression, baseline MMSE, and total energy intake was used to analyze the association of the total AN intake with the risk of dementia.

In order to verify the stability of the model (as the generalization ability of the model on the new dataset), we used the 10-fold cross-validation to verify the model, divided the total participants into 10 subgroups in random, chose one subgroup as the test group to complete the prediction and validation of the established model. The remaining nine subgroups were used as the training group to complete the model establishment process, performed ten times independent cross-validation tests, with each subgroup serving as the test group in sequence, averaged the results of ten times cross-validation tests to obtain a single estimate. The estimation indicators were root mean square error (RMSE) and mean absolute error (MAE). The closer RMSE and MAE to 0, the stronger stability of the model [40].

As for the subgroup analysis, we converted the continuous variables of age and education year to two discrete variables by the median age as 70 (≥70 years/<70 years) and the median education years as 12 (≥12 years/<12 years). Other baseline characteristics, such as gender, APOE-ε4, hypertension, diabetes, depression, and obesity, were two categorical variables (yes/no).

SPSS 25 and R 4.2.1statistical software were used for data analysis. p values and 95% CI were two sided. p < 0.05 was considered as the difference, or association was statistically significant.

A total of 2,967 participants who met the inclusion and exclusion criteria were included in the cohort, and 1,550 participants ultimately completed the follow-up period (online suppl. Fig. S1; for all online suppl. material, see https://doi.org/10.1159/000541231). As shown in Table 1, 1,550 participants with a median age of 70 (interquartile range 65–76) were followed up for an average of 5.2 (standard deviation 0.9) years (i.e., 8,111.3 person-years), and 135 (8.7%) participants converted to dementia.

Based on the restrictive cubic spline shown in Figure 1, we categorized the study participants into two groups: low AN intake group (total AN intake <112 mg/day, n = 380) and high AN intake group (total AN intake ≥112 mg/day, n = 1,170). Participants with high AN intake accounted for 75.5% of the total. Participants with low AN intake had less total energy intake and higher proportions of APOE-ε4 positive, obesity, and depression than those with high AN intake. The cumulative incidence rate of dementia in the low AN intake group was 3.5/100 person-years, which was significantly higher than that in the high AN intake group (1.1/100 person-years, p < 0.001).

As shown in Figure 2 and online supplementary Table S1, daily intakes of carotene (HR: 2.82, 95% CI: 1.69–4.74), vitamin C (HR: 4.52, 95% CI: 2.71–7.54), vitamin E (HR: 4.24, 95% CI: 2.45–7.33), lutein (HR: 3.00, 95% CI: 1.80–5.02), and flavonoid (HR: 2.97, 95% CI: 1.83–4.80) were associated with incident dementia, respectively, adjusting for age, gender, education year, obesity, APOE-ε4, hypertension, diabetes, depression, baseline MMSE, and total energy intake.

The cumulative incidence rate of dementia in participants with low AN intake increased more dramatically than that in those with high AN intake (log-rank test, p < 0.001) (Fig. 3). An 87% higher risk of incident dementia (HR: 1.87, 95% CI: 1.26–2.77) was found in individuals with low AN intake after adjusted by age, gender, education year, obesity, APOE-ε4, hypertension, diabetes, depression, baseline MMSE, and total energy intake. This result was supported by the 10-fold cross-validation with the RMSE = 0.59 and the MAE = 0.72. The model was stable (Table 2).

As shown in Figure 4, the participants with “low AN intake” and ≥70 years, <12 years of education, APOE-ε4 positive, depression, diabetes, hypertension, or obesity had the highest risks of dementia in analysis of subgroup with age ≥70 years (HR: 29.26, 95% CI: 13.39–36.97), subgroup with education <12 years (HR: 3.64, 95% CI: 6.65–18.41), subgroup with APOE-ε4 positive (HR: 5.91, 95% CI: 3.34–10.47), subgroup with depression (HR: 5.12, 95% CI: 2.98–8.82), subgroup with diabetes (HR: 4.67, 95% CI: 2.45–8.92), subgroup with hypertension (HR: 6.31, 95% CI: 3.88–10.26), and subgroup with obesity (HR: 4.84, 95% CI: 2.58–9.07). There was no significant difference in risk of dementia between males and females in those participants with the same AN intake level. In the subgroup with <70 years, no difference with the dementia risk was found between high AN intake and low AN intake groups (HR: 1.88, 95% CI: 0.39–9.07).

In this prospective cohort study, an 87% higher risk of incident dementia was found in individuals with low AN intake (<112 mg/day) compared with those with high AN intake (≥112 mg/day). The additive effects of low AN intake and age ≥70 years, female, education <12 years, APOE-ε4, hypertension, diabetes, depression, and obesity with incident dementia were also demonstrated. This is one of the few prospective studies providing evidence of the association of total dietary AN intake with the risk of severe cognitive impairment among Chinese older adults.

A recent meta-analysis suggested that high intakes of β-carotene, vitamins C and E could decrease the risk of AD, which was consistent with our study. It also showed that the most obvious beneficial associations were observed for vitamin E, followed by β-carotene and vitamin C [41]. In the US MacArthur Research Network Study, β-carotene in particular may offer protection from cognitive decline in people over 65 years. The adjusted odds ratio (OR) of high β-carotene level for cognitive decline was 0.11 (95% CI: 0.02–0.57) in participants with APOE-ε4 positive and 0.89 (95% CI: 0.54–1.47) among those who were APOE-ε4 negative [42]. This finding is consistent with our study on the part of the relation between dietary carotene intake and cognitive impairment.

In the Canadian Study of Health and Aging, older adults with combined use of vitamin C and E supplements and/or multivitamin consumption at baseline were less likely (adjusted OR: 0.51, 95% CI: 0.29–0.90) to have significant cognitive decline during a 5-year follow-up [43]. In the present study, we also observed such associations between dietary vitamin E and C intake and incident dementia. Our result is also consistent with the Memory and Aging Project [44] and the Western Sydney Stroke in the Elderly Study [45]. The Memory and Aging Project was a cohort study of 960 participants aged 58–99 years who cooperated in an FFQ and ≥2 cognitive assessments over a mean of 4.7 years. This study found that consumption of approximately 1 serving per day of α-tocopherol (from 4.3 to 8.1 mg/day) may slow the cognitive decline with aging (β = 0.03, p = 0.020) [38]. The Western Sydney Stroke in the Elderly Study followed a cohort of 179 participants aged 69–91 years for an average of 5 years and found that consumption of vitamin C was associated with a lower prevalence of more severe cognitive impairment (based on the MMSE score, adjusted OR: 0.39, 95% CI: 0.18–0.84) [39]. The French PAQUID study reported that flavonoid intake was related to a better cognitive performance at baseline (p = 0.019) and a better evolution of the performance over time (p = 0.046) [15]. Lutein, the predominant carotenoid in human brain tissue, may play a role in the maintenance of cognition through its anti-inflammation and anti-oxidation properties [46]. A recent meta-analysis of randomized controlled trials found that dietary lutein was associated with slight improvements in cognitive performance in complex attention (standardized mean differences, SMD: 0.02, 95% CI: 0.27–0.31), executive function (SMD: 0.13, 95% CI: 0.26–0.51), and memory (SMD: 0.03, 95% CI: 0.26–0.32), but its effect was not statistically significant [47].

However, the ARIC study [17] and the Doetinchem Cohort Study [48] did not find significant correlations of antioxidant vitamin intake or supplement use (except for lignans) with any cognitive test. There might be several reasons for the inconsistency with our study. On the one hand, the outcome indicators of the ARIC study were cognitive function test scores, while our study took the incidence of dementia as the outcome indicator. On the other hand, the study participants in these 2 studies were middle-aged individuals with the age range of 48–67 and 20–59 years, while our study population was older adults aged ≥60 years. The cognitive impairment of middle-aged individuals may mostly be caused by other diseases rather than neurodegeneration.

Reactive oxygen species (ROS) are generated through the oxidation process by a variety of sources from the environment and normal cellular functions, such as free radicals, non-radical oxygen species and reactive lipids, and carbohydrates [49]. ROS can cause oxidative damage to DNA in routes of the oxidative modification of the nucleotide bases, sugars, or by forming cross-links that lead to mutations, pathologies, cellular aging, and death. Protein oxidation may play a causative role in neurodegenerative diseases such as AD. When the radical meets and reacts with an antioxidant molecule such as vitamin C or E, the reactivity of ROS can be terminated and the oxidative damage to DNA can be reduced [49, 50]. This might be the potential mechanism why lower AN intake is associated with a higher risk of dementia.

Several limitations should be noted in this study. First, during the 5-year follow-up, the changing environment and the new onset of other diseases may influence the association. Second, we assumed that the dietary habits of older people were relatively stable across the follow-up time and therefore defined the baseline dietary intake as the exposure. However, changing in dietary habits might exist, and FFQ method might be unreliable and underreported due to the recall bias. The stability and reliability may need further validation. Third, we simply added up the intake of each individual AN as the total AN intake. The different antioxidant capacities in these individual ANs might impact the accuracy of the cut-off value of the exposure measurements and estimation of the associations in our study. There may be some correlations among ANs; therefore, ANs should be looked individually. The antioxidant capacity of each individual AN should be weighted by a composite score approach in further research. Fourth, we did not consider minerals such as selenium, zinc, manganese, which were also defined as ANs. Our results may be impacted although the intake of minerals by most Chinese people was little according to their diet. Fifth, because of our limited sample size, we did not adjust for the intake of compounds that could potentially influence the bioavailability, absorption, and/or metabolic turnover of AN, especially BMI, which may have determined different risk changes for AN (although we adjusted obesity in the statistical models). Therefore, the estimation of the associations in our study should be explained cautiously. Sixth, 48% of participants were not able to be followed up. As shown in online supplementary Table S2, participants who were lost to follow-up were older, with lower BMI, more likely to have hypertension and depression, with less total energy, individual AN, and total AN, and with lower MMSE at baseline than those who completed the follow-up. Our results may be underestimated because individuals who were lost to follow-up may have a higher risk of dementia. Seventh, we did not have a summary measure of vascular risk as Framingham Risk Score; therefore, our results did not include the contribution of the vascular risk factors, although we adjusted obesities, hypertension, and diabetes in the statistical models. Eighth, we did not consider the total AN intake-dementia association specifically in the subgroup of participants with MCI. Since individuals with MCI may have a higher probability of developing dementia, further studies should be conducted focusing on this special group. Finally, a potential selection bias of our study samples cannot be ignored because our study participants were older adults residing in Shanghai, a metropolis in China. Thus, our results may not be well generalized to other populations.

In conclusion, our study found that low total AN intake may be a risk factor for incident dementia among older adults. Our results suggest that, in the older population, maintaining sufficient AN intake may be beneficial against age-related cognitive decline and dementia. Further prospective studies with larger sample size and longer follow-up should be carried out to verify our findings. Furthermore, multidimensional measurement of AN concentration in the human body, i.e., intake level and blood concentration, should also be considered in the study design. Randomized controlled trial to evaluate the efficacy of individual or combined AN micronutrient dietary supplements should be further implemented to provide the evidence to the overall management of dementia.

The present study was approved by the Medical Ethics Committee of Huashan Hospital at Fudan University (No. 2009-195). Written informed consents were provided by all the participants at the baseline recruitment.

The authors declare no conflict of interest.

This study was supported by the Nutrition and Foods Branch of China Association of Gerontology and Geriatrics, Shanghai Municipal Science and Technology Major Project (2018SHZDZX01) and ZJ LAB, National Natural Science Foundation of China (82173599, 81773513, 82071200), Key Project of the Ministry of Science and Technology, China (2020YFC2005003, 2021YFE0111800), Shanghai Hospital Development Center (SHDC2020CR4007), Shanghai Municipal Health Commission (2020YJZX0101), and MOE Frontiers Center for Brain Science (JIH2642001/028).

Study conception and design and critical revision and commentary on manuscript: J.L., D.D., Y.W., and W.F. Clinical diagnosis and interpretation of the data: Q.Z., Z.X., and D.D. Acquisition, analysis, or interpretation of data: S.L., J.L., W.W., X.Z., S.D., Y.W., and D.D. Statistical analysis: S.L., X.L., and J.L. Manuscript drafting: S.L., J.L., and D.D. All authors read and approved the final manuscript.

Su Liu and Jianfeng Luo contributed equally to this paper.

Data in the current study are available from the corresponding author Prof. Ding Ding on reasonable request and with permission of Huashan Hospital.