Introduction

Low-grade chronic inflammation is a common underlying process for several age-related chronic diseases and conditions [1], amongst others Alzheimer’s disease [2]. At the brain level, neuro-inflammation is highly involved in neuronal damage, which may lead to cognitive decline. Besides, an elevated inflammatory status associated with activated microglia has been observed in demented patients [3–5].

Although pro-inflammatory molecules such as some cytokines are not able to cross the blood–brain barrier, it is well recognized that systemic inflammation and neuro-inflammation are closely related [6], arguing for a role of peripheral inflammation in cognitive aging.

Nevertheless, the role of systemic and neuro-inflammation in the etiology of neurodegenerative diseases is still debated [6].

Epidemiologic research focusing on the relationship between circulating pro-inflammatory markers, including C-reactive protein (CRP), interleukin (IL)-6, IL-10, tumor necrosis factor-a, and cognitive outcomes has been recently reviewed [7, 8]. Data are relatively scarce, and existing findings from longitudinal studies are contrasted, with some studies reporting a positive association [9–15] while others reporting null findings [9, 16–19]. The results also vary across the studied cognitive domains and pro-inflammatory biomarkers tested. Additionally, most studies have been carried out among the elderly, and thus, a potential for elevated inflammatory status induced by preclinical disorders cannot be excluded, thus limiting causal inference.

There is a growing body of evidence suggesting the anti-inflammatory effects of specific dietary factors including n-3 polyunsaturated fatty acids, fibers, vitamins and minerals, as well as overall dietary patterns [20, 21]. This suggests new directions for nutritional prevention via decreasing systemic inflammation by modulating cytokine production.

Recently, based upon existing mechanistic and epidemiological data from about 1943 published studies linking chronic systemic inflammation and nutritional factors, the dietary inflammatory index (DII) reflecting the overall inflammatory potential of the diet has been developed and validated [22, 23]. Unlike considering individual food or nutrients, the use of the DII may provide complementary arguments for the role of the dietary inflammatory potential in health. Furthermore, the DII has been associated with pro-inflammatory markers in some [23–27], but not all [28] epidemiological studies to date.

The purpose of the present study was to examine the long-term association between the DII evaluated in midlife using repeated 24-h records and overall and domain-specific cognitive functioning measured 13 years later, using a large cohort. Specifically, we hypothesized that long-term healthy dietary habits may help to maintain cognitive functioning during aging through their anti-inflammatory properties, expressed as lower DII scores.

Materials and methods

Population

The French SU.VI.MAX «SUpplémentation en Vitamines et Minéraux AntioXydants» study (1994–2002) was a randomized, double-blind, placebo-controlled, primary prevention trial initially designed to test the potential efficacy of daily supplementation with antioxidant vitamins and minerals at nutritional doses on the incidence of cancer, ischemic heart disease, and overall mortality. It included 12,741 adult volunteers living in France (women aged 35–60 years and men aged 45–60 years). The initial follow-up was planned for 8 years [29]. Five years after the end of the trial phase, a total of 6850 individuals were included in the SU.VI.MAX 2 observational follow-up study (2007–2009). The SU.VI.MAX and SU.VI.MAX 2 studies were conducted according to the Declaration of Helsinki guidelines. All procedures involving human subjects were approved by the Ethics Committee for Studies with Human Participants of Paris-Cochin Hospital (CCPPRB No. 706 and No. 2364, respectively) and the Commission Nationale de l’Informatique et des Libertés (CNIL No. 334641 and No. 907094, respectively). Written informed consent was obtained from all subjects. The SU.VI.MAX trial was registered at www.clinicaltrials.gov under #NCT00272428.

Dietary data assessment

During the SU.VI.MAX trial, dietary data were collected through 24-h dietary records provided via computerized questionnaires using the Minitel, a small terminal widely used in France in the 1990s as an adjunct to the telephone. A maximum of six randomly distributed records was completed per year, covering weekdays and weekend days and all seasons of the year. Participants were assisted by an instruction manual that included validated photographs of >250 generic foods shown in three main portion sizes [30]. The Phenol-Explorer database [31] and a published validated composition table [32] were used to compute nutrient intakes. For the present analysis, the food and nutrient intakes were calculated as the mean reported values across all eligible 24-h records collected during the first 2 years of the SU.VI.MAX study. Next, the DII was computed using a previously published algorithm [22]. Briefly, individual intake of nutrients and food was standardized using worldwide mean and standard deviation (SD) values. To prevent skewing, z-scores were converted to centered percentiles which were then multiplied by an effect score. It was based on a literature review of 1943 studies examining the relationship between dietary constituents and inflammatory markers, to obtain a food-specific value. At the individual level, all such values were summed up to obtain the overall DII. In the present study, the DII includes data on 35 of the 45 food parameters: carbohydrates, protein, total fat, alcohol, fiber, cholesterol, saturated fatty acids, monounsaturated fatty acids, polyunsaturated fatty acids, omega 3, omega 6, niacin, thiamin, riboflavin, vitamin B6, vitamin B12, iron, zinc, vitamin A, vitamin C, vitamin D, vitamin E, folic acid, beta carotene, anthocyanidins, flavan-3-ol, flavonols, flavonones, flavones, isoflavones, garlic, ginger, pepper, onion, and tea.

Cognitive assessment

During the SU.VI.MAX 2 study (2007–2009), all participants were invited to undergo a neuropsychological evaluation. Episodic memory was evaluated with the “rappel indicé-48” (RI-48), a validated cued recall test (list of 48 words from 12 categories). For each word correctly retrieved, one point was attributed (theoretical maximal score = 48). Lexical–semantic memory was assessed using a phonemic fluency task (citing words beginning with the letter P) and a semantic fluency task (naming as many animals as possible). For each task, the score was the number of words correctly produced during a 2-min period. Forward and backward digit span tasks (issued from the validated Wechsler adult intelligence scale-third edition [33]) were used to assess short-term and working memory. These tasks consisted in immediately repeating an increasing sequence of digits, with one point attributed for each sequence repeated correctly until the participant failed two consecutive trials of the same digit span (theoretical maximal score for each task =14). Mental flexibility was assessed with the Delis–Kaplan trail-making test (TMT) which is a number–letter switching task. The score was the time, in seconds, needed to complete task. The inverse of the TMT score was log-transformed to improve normality. We converted the initial cognitive test scores into T scores (mean = 50, SD = 10), so that a one-point difference in the test score corresponded to one-tenth of an SD difference and each score was standardized, thus allowing for comparison.

A composite cognitive score was extracted by principal component analysis (SAS Proc Factor). The factor accounted for 41 % of the initial variance. The factor loadings (i.e., correlation coefficients between the composite cognitive score and each individual neuropsychological tests) are given in Supplemental Table 1).

Covariates

Upon enrollment, information on sex, date of birth, smoking status (never-smoked, former, or current smoker), physical activity (irregular, <1 h walking/day or equivalent, ≥1 h walking/day or equivalent), formal education (primary, secondary, or post-secondary), occupational category (homemaker, farmer/manual labor, artisan/self-employed/office worker/skilled labor, managerial staff/intellectual profession), and perceived memory troubles (yes/no) was collected via self-administered questionnaires.

Anthropometric and clinical measurements (including body mass index (BMI kg/m2) and blood pressure) were obtained at baseline and at the end of the follow-up, as previously described [34]. Information on hypertension (≥140/90 mmHg or medication use) and diabetes (fasting blood glucose ≥7 mmol/l, or anti-diabetic medication use) over the follow-up was also collected as previously reported [34]. During the follow-up, all reported cardiovascular events were validated by an independent expert committee. Depressive symptoms were assessed at follow-up (2007–2009) using the French version of the center for epidemiologic studies depression scale (CES-D), and the total score was used as a covariate [35].

Statistical analysis

In the present analysis, we included subjects aged 45–60 years at baseline with available cognitive evaluation (N = 4447) and dietary data (i.e., ≥3 24-h records over the first 2 years of follow-up; N = 3362). Subjects with missing values for any of the covariates were excluded, leaving a final sample of 3080 participants.

We compared included and excluded subjects using Chi-square tests or Wilcoxon’s rank tests. Baseline characteristics are presented by sex-specific quartiles (Q) of the DII. Values represent mean (±SD) or percentages, and P values were calculated using linear contrast or Chi-square trend tests, as appropriate. Cross-time associations between the DII at midlife and cognitive function 13 years later were analyzed by ANCOVA. The first model was adjusted for age at the neuropsychological evaluation and sex. The second model was further adjusted for education, follow-up time between baseline and cognitive evaluation, supplementation group during the trial phase, number of 24-h dietary records, energy intake, BMI, occupational status, tobacco use status, physical activity, memory troubles, depressive symptoms concomitant with the cognitive function assessment, and history of diabetes/hypertension/cardiovascular disease.

All analyses were conducted with SAS® software (version 9.3, SAS Institute, Cary, NC, USA) with a significance level of 0.05 (two-sided tests).

Sensitivity analyses

As dietary intakes were estimated from at least three 24-h records, this might potentially lead to inability to control for intra-individual variability. In order to test the robustness of the main findings, we reanalyzed our data among participants with at least six 24-h records (N = 2695).

Results

Participant characteristics

Compared with those included, excluded participants were younger, more often female, smokers, with lower levels of physical activity and education. They also had lower energy intake, lower CES-D scores and exhibited higher (i.e., more pro-inflammatory) DII values (data not shown).

At baseline, the mean age of our study sample was 52.0 ± 4.6 years, and at the time of the cognitive evaluation, the mean age was 65.4 ± 4.6 years. The mean (SD) DII was 0.41 (1.84), and it was lower in men than in women [0.34 (1.80) and 0.97 (1.88), respectively].

Baseline characteristics across Q of the DII are presented in Table 1. Compared with those with a low DII scores, participants with a high DII scores, reflecting a more pro-inflammatory diet, were less educated, less physically active, and more often smokers.

Table 1 Baseline characteristics of the sample across quartiles (Q) of the dietary inflammatory index, SU.VI.MAX study (n = 3080)
Full size table

The DII was negatively correlated with energy intake and percent of energy from carbohydrates and positively correlated with percent of energy from lipids and proteins. After adjustment for total energy, the DII was negatively correlated with folic acid, beta carotene, vitamin C, PUFA, and fiber intakes.

Association between the DII and global cognitive functioning

Table 2 presents the association between the DII modeled in Q and global cognitive functioning assessed through principal component analysis. A higher DII, reflecting a pro-inflammatory diet, was associated with lower subsequent global cognitive function [mean difference Q4–Q1 = −1.76 (−2.81; −0.72), P for trend =0.002 in the fully adjusted model].

Table 2 Association between the dietary inflammatory index modeled in quartiles (Q) and global cognitive function, SU.VI.MAX study (n = 3080)
Full size table

Association between DII and individual cognitive tests

Table 3 presents the association between the DII in Q and performance on each individual cognitive test. For the RI-48 cued recall test and semantic and phonemic fluencies, a higher DII score was associated with a poorer performance.

Table 3 Association between the dietary inflammatory index modeled in quartiles (Q) and cognitive performance by domain, SU.VI.MAX study (n = 3080)
Full size table

In addition, in fully adjusted models, no association between the DII and executive functioning-related cognitive tests (the TMT, forward and backward digit span) was observed.

For comparison purposes, the associations between the DII modeled as a continuous variable and individual cognitive tests scores are shown in Fig. 1. The mean differences in cognitive performances for a 1 unit change in the DII score are shown.

Fig. 1
figure 1

Association between the dietary inflammatory index as a continuous variable and cognitive test scores. Values are adjusted mean difference (95 % confidence interval) in cognitive test scores (estimated via ANCOVA) using the first quartile of the dietary inflammatory index as reference. Adjustment is made for age, gender, education, follow-up time between baseline and cognitive evaluation, supplementation group during the trial phase and number of 24-h dietary records, energy intake, BMI, occupational status, tobacco use status, physical activity, memory troubles at baseline, depressive symptoms concomitant with the cognitive function assessment, history of diabetes/hypertension/cardiovascular disease

Full size image

Sensitivity analyses

We repeated our analyses among participants with at least six 24-h records thus improving the estimation of nutritional intakes. Findings were similar in terms of direction, though the associations were strengthened (data not shown).

Discussion

In this French study, the DII estimated at midlife was inversely associated with subsequent cognitive functioning. The association was strong even after adjustment for a wide range of potential confounders, including sociodemographic, lifestyle, and health factors, in particular vascular conditions. This link was mostly driven by the associations with cognitive tests reflecting verbal memory. No association was observed with executive functioning assessed via digit span tasks and the TMT.

Inflammation and cognitive function

The few prospective cohort studies investigating the role of inflammation in later cognitive outcomes have not provided consistent findings. Methodological flaws have been advanced, such as reverse causality (due to an existing preclinical stage) and the use of a single biomarker for the assessment of inflammation, as explanations of this phenomenon [7, 8]. To the best of our knowledge, only two studies were carried out among relatively young individuals, where reverse causality was unlikely [10, 12]. In the Honolulu–Asia aging study, cognitive decline was positively associated with higher CRP concentrations in midlife, but no association was detected after removing dementia cases occurring during the follow-up [10].

In the Whitehall study [12], the authors reported a predictive role of blood concentrations of the inflammatory biomarker IL-6, but not CRP in cognitive decline (overall function and reasoning). These studies and our findings provide new insights and extend knowledge concerning the association between inflammatory status in midlife and later global cognitive health during aging. Our results support the hypothesis that a pro-inflammatory diet may impact both cognitive functioning as a whole, and regarding specific cognitive domains, by contributing to a state of systemic inflammation.

In our study, we observed a strong inverse association between a pro-inflammatory diet and cognitive performance on the fluency tests. However, further investigation with respect to the identification of cognitive domains susceptible to inflammation is needed. Different authors have reported associations with global cognitive functioning [9, 11–13], reasoning [12], motor speed [14], executive function [15], and memory function [15]. However, data are sparse and further research is needed to elucidate the differential susceptibility of specific cognitive domains to inflammation.

We identified three studies reporting significant associations between IL-6 levels and cognitive outcomes, but no association with CRP [9, 14, 36], suggesting that early inflammatory biomarkers may prospectively predict future cognitive outcomes. Moreover, these studies argue for the need to focus on several inflammatory biomarkers.

While systemic inflammation is the product of many factors including diet, obesity, and other conditions, the present study aimed to specifically focus on the part of systemic inflammation that can be attributed to nutritional factors. Although it can be assumed that dietary factors have a smaller effect on inflammation than other factors such as age and genetic factors, they are important in terms of public health since they are potentially modifiable. The scientific literature linking dietary patterns and inflammation is growing [20, 37], and the DII has been previously shown to be associated with inflammatory markers [23–27, 38]. Strong findings were obtained from the SEASONS study, a longitudinal follow-up study of 550 adult men and women in whom the DII clearly predicted interval changes in blood levels of CRP [23]. All of these studies argue for an association between DII and clinical markers of inflammation. This diet-based approach can give important new insights into the long-term preventive potential of a healthy diet, as a complement to studies focusing on single nutrients (which allows to obtain specific information on particular nutrients) or on single foods (which allows to also consider food matrix effects).

Diet quality index and cognitive function

Our findings are consistent with previous research focusing specifically on the role of vitamin D, a variety of polyphenols, beta carotene, and vitamin C and vitamin E intakes on cognitive functioning [39–41].

Besides, as expected, the DII was correlated with a less healthy nutritional profile [42]. Hence, our findings can be interpreted in light of previous research focusing on dietary quality using holistic approaches and the link with cognitive outcomes. Indeed, the present findings are consistent with our previous work documenting better cognitive functioning among subjects with a higher level of adherence to nutritional guidelines at midlife [43]. The findings are also consistent with those based on a priori dietary indexes, in particular the Mediterranean diet score, and a posteriori data-driven approaches such as factor analysis [44, 45]. In particular, the Mediterranean diet, which has been hypothesized to exhibit anti-inflammatory properties and beneficial vascular effects [46, 47], has been consistently associated with several cognitive outcomes in a recent meta-analysis of five longitudinal studies [48].

Strengths and limitations

Our findings should be interpreted taking into account some limitations. First, we were not able to focus on cognitive decline as cognitive performance was not measured at baseline. However, due to the relatively young age of this population heavily involved in a cohort study, it is likely that the completion of many questionnaires over a long follow-up period (13 years) argues against prevalent cognitive impairment at baseline inducing bias in any of the measures, including dietary intake. Finally, as participants of the SU.VI.MAX study were volunteers, the external validity of our findings might be limited. Concerning the use of the DII, similarly to other a priori indexes, it might be subject to some limitations, including the arbitrary selection of components and scoring methods [49]. In addition, in our study, the DII was constructed using 35 parameters, while the original index includes 45 food parameters. This may have led to some misestimation of the dietary inflammatory potential.

Our study also exhibits strengths and important original aspects, including a large cohort of community-dwelling subjects, a focus on midlife exposures, and the use of relatively accurate dietary data. Lastly, we use neuropsychological standardized tests with good sensitivity, and avoiding ceiling and floor effects.

Conclusion

In conclusion, our study supports a significant harmful role of a pro-inflammatory diet in midlife in subsequent cognitive function. This suggests that diet may act on cognitive functioning through its inflammation-inducing properties. The findings can help to refine population-level guidelines regarding the maintenance of cognitive health during aging.