Hostname: page-component-7c8c6479df-94d59 Total loading time: 0 Render date: 2024-03-17T13:57:04.830Z Has data issue: false hasContentIssue false

Carotenoid-rich dietary patterns during midlife and subsequent cognitive function

Published online by Cambridge University Press:  27 September 2013

Emmanuelle Kesse-Guyot*
Affiliation:
Université Paris 13, Sorbonne Paris Cité Université, UREN (Nutritional Epidemiology Research Unit), Inserm (U557), Inra (U1125), Cnam, SMBH Paris 13, 74 rue Marcel Cachin, 93017Bobigny Cedex, France
Valentina A. Andreeva
Affiliation:
Université Paris 13, Sorbonne Paris Cité Université, UREN (Nutritional Epidemiology Research Unit), Inserm (U557), Inra (U1125), Cnam, SMBH Paris 13, 74 rue Marcel Cachin, 93017Bobigny Cedex, France
Véronique Ducros
Affiliation:
Département de Biologie Intégrée, CHU de Grenoble, Grenoble, France
Claude Jeandel
Affiliation:
Centre de Gérontologie, Clinique Antonin Balmes, CHU Montpellier, Université I, Montpellier, France
Chantal Julia
Affiliation:
Université Paris 13, Sorbonne Paris Cité Université, UREN (Nutritional Epidemiology Research Unit), Inserm (U557), Inra (U1125), Cnam, SMBH Paris 13, 74 rue Marcel Cachin, 93017Bobigny Cedex, France Département de Santé Publique, Hôpital Avicenne (AP-HP), Bobigny, France
Serge Hercberg
Affiliation:
Université Paris 13, Sorbonne Paris Cité Université, UREN (Nutritional Epidemiology Research Unit), Inserm (U557), Inra (U1125), Cnam, SMBH Paris 13, 74 rue Marcel Cachin, 93017Bobigny Cedex, France Département de Santé Publique, Hôpital Avicenne (AP-HP), Bobigny, France
Pilar Galan
Affiliation:
Université Paris 13, Sorbonne Paris Cité Université, UREN (Nutritional Epidemiology Research Unit), Inserm (U557), Inra (U1125), Cnam, SMBH Paris 13, 74 rue Marcel Cachin, 93017Bobigny Cedex, France
*
*Corresponding author: E. Kesse-Guyot, fax +33 1 48 38 89 31, email e.kesse@uren.smbh.univ-paris13.fr
Rights & Permissions [Opens in a new window]

Abstract

Carotenoids may help to prevent the ageing of the brain. Previous findings regarding β-carotene alone are not consistent. In the present study, we evaluated the cross-time association between a carotenoid-rich dietary pattern (CDP) and subsequent cognitive performance using a sample of 2983 middle-aged adults participating in the SU.VI.MAX (Supplémentation en Vitamines et Minéraux Antioxydants) study. Cognitive performance was assessed in 2007–9 using six neuropsychological tests, and a composite cognitive score was computed. The cognitive data were related to dietary data obtained by repeated 24 h dietary records (1994–6) and to measurements of baseline plasma concentrations of carotenoids (lutein, zeaxanthin, β-cryptoxanthin, lycopene, α-carotene, trans-β-carotene and cis-β-carotene). DP were extracted using the reduced rank regression method for 381 participants and then extrapolated to the whole sample using plasma carotenoid concentrations as response variables. Associations between a CDP and cognitive function measured 13 years later were estimated with ANCOVA providing mean difference values and 95 % CI across the tertiles of CDP. A correlation between CDP and consumption of orange- and green-coloured fruits and vegetables, vegetable oils and soup was observed. CDP was found to be associated with a higher composite cognitive score (mean difference 1·04, 95 % CI 0·20, 1·87, P for trend 0·02), after adjustment for sociodemographic, lifestyle and health factors. Similar findings were obtained for scores obtained in the cued recall task, backward digit span task, trail making test and semantic fluency task (all P for trend < 0·05). Further studies ought to confirm whether a diet providing sufficient quantity and variety of coloured fruits and vegetables may contribute to the preservation of cognitive function during ageing.

Type
Full Papers
Copyright
Copyright © The Authors 2013 

With the increasing average lifespan of humans, the prevalence of age-related cognitive decline is rising. No treatments are available to cure or slow down cognitive decline, which makes prevention a critical strategy to address age-related cognitive disorders( Reference de la Torre 1 ). Nutritional factors, being modifiable, elicit due interest in the prevention of age-related cognitive decline, and solid understanding of their potential influence could help to identify targets for intervention. However, the low quantity of evidence from the available scientific literature( Reference Middleton and Yaffe 2 Reference Plassman, Williams and Burke 5 ) suggests that further studies are needed.

Carotenoids are natural pigments present in plant-based foods and are well known for their role as efficient scavengers of reactive oxygen species and may thus help to protect the brain against oxidation occurring during the ageing process( Reference Gemma, Vila, Bachstetter and Riddle 6 , Reference Polidori, De Spirt and Stahl 7 ). Indeed, the brain is especially prone to oxidative stress owing to its high content of long-chain PUFA that are extremely sensitive to peroxidation and to increased production of reactive oxygen species due to a high metabolic rate( Reference Gemma, Vila, Bachstetter and Riddle 6 ). Reactive oxygen species are essential in signalling pathways; however, any increase in reactive oxygen species production may be detrimental to lipids, proteins and DNA (especially at the mitochondrial level) as a result of the increase in oxidative stress and dysfunction in signal cascades and/or apoptosis( Reference Federico, Cardaioli and Da 8 Reference Patten, Germain and Kelly 10 ). Carotenoids also exhibit anti-inflammatory properties probably through the modulation of the lipoxygenase enzyme and activation of the expression of genes involved in cell communication( Reference Donaldson 11 , Reference Giordano, Scicchitano and Locorotondo 12 ).

Most of the available epidemiological literature on carotenoid-based prevention of brain ageing has been focused on β-carotene alone, in spite of other carotenoids being found to have interesting properties( Reference Polidori, De Spirt and Stahl 7 ).

Cross-sectional and case–control studies have indeed reported associations between lower intake or plasma status of β-carotene and lower cognitive function( Reference Jama, Launer and Witteman 13 Reference Ortega, Requejo and Andres 15 ) or higher risk of dementia( Reference Rinaldi, Polidori and Metastasio 16 Reference Wang, Shinto and Connor 18 ). Moreover, longitudinal studies have reported lower cognitive decline( Reference Wengreen, Munger and Corcoran 19 , Reference Hu, Bretsky and Crimmins 20 ) or lower risk of Alzheimer's disease( Reference Engelhart, Geerlings and Ruitenberg 21 ) among subjects with higher intake or plasma status of β-carotene. However, two cross-sectional studies on cognitive function( Reference Schmidt, Hayn and Reinhart 22 , Reference Kalmijn, Feskens and Launer 23 ) and five longitudinal studies on cognitive decline( Reference Kang and Grodstein 24 , Reference McNeill, Jia and Whalley 25 ) and risk of dementia( Reference Laurin, Masaki and Foley 26 Reference Devore, Grodstein and van Rooij 28 ) have not reported such associations.

Besides, randomised controlled trials have not yet confirmed any potential benefits of β-carotene supplementation alone or in combination with other antioxidants regarding cognitive decline or dementia prevention( Reference Kang, Cook and Manson 29 Reference Yaffe, Clemons and McBee 31 ), except for one such trial carried out among participants in the Physicians' Health Study II. In that study, short-term supplementation with a low dose of β-carotene was found to be not associated with cognitive function, while a beneficial impact of long-term supplementation was observed( Reference Grodstein, Kang and Glynn 32 ).

Prior research exploring the relationship between other carotenoids and cognitive ageing is scarce, with study designs and findings being inconsistent( Reference Akbaraly, Faure and Gourlet 14 , Reference Wang, Shinto and Connor 18 , Reference Schmidt, Hayn and Reinhart 22 , Reference Kang and Grodstein 24 , Reference Johnson 33 , Reference Polidori, Mattioli and Aldred 34 ). In a cross-sectional study, an association between cognitive impairment and low plasma lycopene and zeaxanthin concentrations has been reported( Reference Akbaraly, Faure and Gourlet 14 ).

Using data from the SU.VI.MAX (Supplémentation en Vitamines et Minéraux Antioxydants) 2 study, the present study aimed to evaluate the association between dietary patterns (DP) that maximally account for the variation in plasma carotenoid concentrations and subsequent cognitive performance, employing the recently proposed reduced rank regression (RRR) statistical technique( Reference Hoffmann, Schulze and Schienkiewitz 35 ).

Materials and methods

Study population

The SU.VI.MAX study (1994–2002; n 12 741) was a randomised, double-blind, placebo-controlled, primary prevention trial evaluating the effect of daily low-dose antioxidant supplementation on the incidence of cancer and IHD( Reference Hercberg, Galan and Preziosi 36 , Reference Hercberg, Preziosi and Briancon 37 ). At the end of the supplementation period (2002), a total of 6850 subjects agreed to participate in a post-supplementation follow-up study (SU.VI.MAX 2, 2007–9).

The SU.VI.MAX and SU.VI.MAX 2 studies were conducted according to the guidelines of the Declaration of Helsinki and were approved by the Ethics Committee for Studies with Human Subjects of Paris-Cochin Hospital (CCPPRB no. 706 and no. 2364, respectively) and the Comité National Informatique et Liberté (CNIL no. 334 641 and no. 907094, respectively). Written informed consent was obtained from all the participants. Clinical trial registration: clinicaltrial.gov (number NCT00272428).

Measurement of carotenoid concentrations

Fasting blood samples were collected in evacuated tubes (Becton Dickinson) at baseline (1994–6). All biochemical measurements were carried out in a single laboratory. Blood samples were centrifuged immediately, frozen and stored at − 80°C. Carotenoids were assessed in the same laboratory using a Biotek-Kontron HPLC system (UVK Lab), which consisted of a 525 dual pump, a 465 autosampler and a 540 diode array detector. The plasma concentrations of β-carotene were measured by Fluka (Sigma-France), and those of other carotenoids were measured by Hoffmann-La Roche (Hoffmann-La Roche). The liquid chromatography separation was carried out with an Alltech Adsorbosphere C18 column (150 × 4·5 mm inner diameter and 3 μm particle size; Templemars), which was thermostated at 28°C with a 402 column oven. Carotenoids were obtained after two extractions with a hexane–tetrahydrofuran mixture. For quantification, we used the method of Steghens et al. ( Reference Steghens, van Kappel and Riboli 38 ) with minor modifications. Specifically, we used a single (instead of two) 150 mm-long column, and we added 10 parts per million water in mobile phase A to improve the separation of retinol, lutein and zeaxanthin.

The limits of detection were calculated as 5-fold the maximum baseline noise in the region of the peaks. Thus, we found limits of detection of 0·02 μm for carotenoids. All the concentrations of retinol, lycopene and β-carotene were above the respective limits, and only 5 % of lutein, 8 % of zeaxanthin and 2 % of β-cryptoxanthin were below these limits.

Dietary data assessment

During the SU.VI.MAX study, the subjects were asked to provide a 24 h dietary record every 2 months via computerised questionnaires. The participants were assisted by an instruction manual that included validated photographs of more than 250 generic foods shown in three main portion sizes( Reference Le Moullec, Deheeger and Preziosi 39 ). A French food composition table was used to estimate nutrient intake( Reference Hercberg 40 ).

Cognitive assessment

During the SU.VI.MAX 2 study (2007–9), all the participants were invited to undergo a medical check-up. This included a clinical examination and a neuropsychological evaluation carried out by trained neuropsychologists. Episodic memory was evaluated with the RI-48 cued recall test (a list of forty-eight words belonging to twelve categories). The score was the number of words retrieved (maximum score of 48)( Reference Ivanoiu, Adam and Van der Linden 41 ). Lexical–semantic memory was assessed by verbal fluency tasks including a semantic fluency task consisting of naming as many animals as possible and a phonemic fluency task consisting of citing words beginning with the letter P. The score was the number of correct words produced during a 2 min period for each task( Reference Lezak, Howieson and Loring 42 ). Short-term/working memory was assessed with the forward and backward digit spans. The subjects were asked to repeat two sequences of digits, forwards and backwards. The number of digits increased by one until the participants failed in two consecutive trials of the same digit span. For each correct sequence repeated, one point was assigned, with a maximum score of 14 for a forward as well as a backward digit span( Reference Wechsler 43 ). Mental flexibility was assessed with the Delis–Kaplan trail making test (TMT) consisting of connecting numbers and letters alternating between the two series. The score was the time in seconds needed to complete a task( Reference Delis, Kaplan and Kramer 44 ), implying that a lower value indicated better performance.

Covariates

Data pertaining to sex, date of birth, occupational category (unemployed, manual worker or blue- and white-collar worker), smoking status (never, former or current), physical activity (irregular, < 1 h walking/d or ≥ 1 h walking/d), education (primary, secondary or post-secondary) and medication use were collected at baseline. In the SU.VI.MAX 2 study, medication use was self-reported.

Anthropometric and clinical measurements (including BMI and blood pressure) were obtained at baseline and at the end of follow-up as described previously( Reference Kesse-Guyot, Andreeva and Jeandel 45 ). Hypertension during follow-up was defined as blood pressure ≥ 140/90 mmHg at any follow-up examination or antihypertensive medication use.

Fasting blood glucose concentrations were measured using an enzymatic method (Advia 1650; Bayer Diagnostics), and diabetes during follow-up was defined as fasting blood glucose concentrations ≥ 7 mmol/l at any follow-up test or anti-diabetic medication use. Depressive symptoms were assessed at follow-up using the French version of the Center for Epidemiologic Studies Depression Scale, and the total score was used as a covariate( Reference Radloff 46 ). Data pertaining to self-reported memory problems were collected at baseline. During follow-up, all the reported cardiovascular events were validated by an independent expert committee.

Statistical analyses

For the present analyses, we selected subjects aged 45–60 years at baseline with available cognitive evaluation data (n 4447) and dietary data (i.e. ≥ 3 24 h dietary records over the first 2 years of follow-up; n 3362). Subjects with missing values for any of the covariates were excluded, leaving a subsample of 2983 participants.

In the present study, the 24 h dietary records provided during the first 2 years of follow-up were taken into account to compute the means of food and nutrient intakes. DP were extracted using the RRR technique described by Hoffman et al. ( Reference Hoffmann, Schulze and Schienkiewitz 35 ). The RRR method is used to derive coherent patterns from a number of predictor variables, thus maximising the explained variation in a pre-defined set of response variables, i.e. intermediate markers that are hypothesised to be related to the health outcome of interest. In the present analyses, responses chosen for the RRR method were plasma carotenoid concentrations available for a subsample of 381 participants.

Participants with and without data on plasma carotenoid concentrations were compared using the Kruskal–Wallis or χ2 test. We first carried out log transformation of plasma carotenoid concentrations to improve normality. The RRR method can extract as many DP as there are response variables. In the present study, all the subsequent models pertain to the first DP, as it reflects the main part of the variation in plasma carotenoid concentrations. Next, an extrapolated DP score was computed for the entire sample. To test the stability and generalisability of the pattern, we applied the bootstrap method with 1000 replications using the sample with plasma carotenoid concentrations. Thus, we estimated the distribution of the Spearman correlation coefficients between these 1000 first RRR-extracted DP and the principal carotenoid-rich dietary pattern (CDP, described below) used in the present analyses. The mean Spearman correlation coefficient was 0·890 (95 % CI 0·889, 0·893). The consumption of a total of thirty different food groups was used as the predictor, which was energy adjusted using the residual method( Reference Willett 47 ). The grouping up of the various foods is described in Supplementary Table S1 (available online).

For the present analyses, we used the inverse of the TMT score, thus a higher score corresponded to a better result. The inverse TMT score was log-transformed to improve normality. Cognitive test scores were converted into T scores (mean 50 (sd 10)). Thus, a 1-point difference in the test score corresponded to 1/10 of a sd difference. A composite cognitive score defined as the mean of the standardised test scores was rescaled to sd= 10.

Descriptive baseline characteristics are reported as means and standard deviations or percentages across tertiles of the extracted DP. The reported P values refer to the Kruskal–Wallis or χ2 trend test, as appropriate. ANCOVA were used to estimate the cross-time associations between the tertiles of the retained baseline DP and the subsequent cognitive performance scores. P values for linear contrast across the tertiles are reported. Mean differences in cognitive scores according to DP score modelled as continuous variables were also estimated using multivariable linear regression. In the initial model, the analyses were unadjusted. In the second set of models, the analyses were adjusted for follow-up time between baseline and cognitive evaluation (continuous variable), sex, supplementation group during the trial phase (active/placebo), education, baseline occupational status, age at cognitive evaluation (continuous variable), intervention group during the SU.VI.MAX trial phase (1994–2002), baseline energy intake (continuous variable), smoking, physical activity, number of available 24 h dietary records, baseline self-reported memory troubles, baseline BMI (continuous variable), depressive symptoms at follow-up (continuous variable), CVD incidence, hypertension and diabetes during follow-up. Further adjustment for the regular use of antioxidant supplements at cognitive evaluation did not have an impact on the estimations; thus, this covariable was not retained in the final model.

To test the robustness of the primary findings, three sets of supplementary analyses were carried out. First, for purposes of partly correcting for selection bias, additional analyses were carried out using inverse probability weighting( Reference Cole and Hernan 48 , Reference Seaman and White 49 ). The probability to be included in the present study was determined for each participant using data on baseline characteristics. The data were reanalysed using the inverse probability to be included as the respective weight. Second, a simplified DP as described by Schulze et al. ( Reference Schulze, Hoffmann and Kroke 50 ) was estimated and further tested with regard to subsequent cognitive function. For this purpose, food group consumption displaying a |correlation coefficient| >0·2 within the DP was retained and standardised (mean 0 (sd 1)) before being summed up. Finally, analyses were carried out only for the subgroup receiving placebo during the trial phase.

Additionally, effect modifications by sex and supplementation group during the trial phase were tested.

Statistical tests were two sided (type 1 error < 0·05). All the analyses were carried out using SAS (release 9.2; SAS Institute, Inc.).

Results

The preliminary analyses showed that participants with and without data on plasma carotenoid concentrations were similar in terms of sociodemographic factors, in particular, sex and education, cardiovascular risk factors (smoking, diabetes, hypertension and BMI) and food consumption (data not shown). However, the subsample with carotenoid status data was slightly older than the other group: 53·6 v. 51·8 years (P< 0·0001).

The retained DP factor, extracted using the RRR method with data from 381 participants with available plasma carotenoid status data, explained 11·66 % of the total variation in the response variables, i.e. plasma carotenoid concentrations, and about 6·46 % of the variation in food group consumption patterns (Table 1). This factor was positively correlated with all the carotenoids, with the highest correlations being evident for carotenes and β-cryptoxanthin. Hence, that factor was termed ‘CDP’.

Table 1 Explained variation in the consumption of foods and in plasma carotenoid concentrations with the carotenoid-rich dietary pattern (n 381)

* Plasma concentrations.

Correlations between the CDP score and the various food groups are presented in Table 2.

Table 2 Food groups associated with carotenoid-rich dietary patterns (extracted from n 2983)

The CDP score was positively correlated with the consumption of green-coloured fruits and vegetables, vegetable oils, orange-coloured fruits and vegetables and soup and was negatively correlated with that of beer, cider and wine.

The mean follow-up duration was 13·6 (sd 0·5) years. The mean age of the population at the time of cognitive evaluation was 65·5 (sd 4·5) years. The characteristics of the participants are given according to tertiles of the CDP score in Table 3. The CDP score was associated with a higher probability of being female and non-smoker, having more formal education and lower BMI, being a blue-collar worker, and being more likely to report the presence of depressive symptoms. It was also negatively associated with alcohol, lipid and protein intake and positively associated with carbohydrate intake.

Table 3 Baseline characteristics of the population across the tertiles of the reduced rank regression-extracted carotenoid-rich dietary pattern score (n 2983)* (Mean values and standard deviations)

CES-D, Center for Epidemiologic Studies Depression Scale.

* Except when otherwise specified.

P values based on non-parametric Kruskal–Wallis test or Mantel–Haenszel χ2 trend test.

Values are presented as the percentage of total daily energy intake (without alcohol).

§ Among n 381 participants.

No interaction between the CDP score and sex or supplementation group regarding cognitive performance was detected (all P values >0·20).

The results of the analyses of the association between the CDP score and cognitive function are given in Table 4. In the unadjusted model, participants with a higher CDP score had higher composite cognitive scores as well as individual scores on the RI-48 cued recall task, backward digit span task, TMT and semantic fluency task. On the other hand, no association was observed for the forward digit span or phonemic fluency task. In the fully adjusted models, these relationships remained statistically significant, despite attenuation of the estimates in most cases.

Table 4 Associations between the carotenoid-rich dietary pattern score (in tertiles (T) and as continuous variable) and cognitive performance (Mean differences and 95 % confidence intervals)

* Values are mean differences and 95 % CI in cognitive test scores (ANCOVA) using the 1st (top) tertile of the carotenoid-rich dietary pattern score as reference.

P for linear contrast.

Model 1 is unadjusted.

§ Model 2 is adjusted for age, sex, 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, reported memory problems at baseline, depressive symptoms concomitant with cognitive function assessment, and history of diabetes/hypertension/CVD.

The findings of the supplementary analyses (inverse probability weighting and the simplified DP) are presented in Supplementary Tables S2 and S3 (available online), respectively. After accounting for potential selection bias using inverse probability weighting, the estimates were slightly attenuated, but no association that was previously significant became non-significant or vice versa. In turn, modelling a simplified DP did not substantially modify the primary findings, except that the CDP and TMT performance were no longer associated. Finally, despite a loss of power, findings obtained for the placebo group were similar to those of the main analysis (data not shown).

Discussion

In the present study, we used a statistical method termed RRR to extract a CDP correlated with the plasma status of various carotenoids. The DP extracted from our data was most strongly correlated with the plasma status of β-carotene, α-carotene, β-cryptoxanthin and lutein. This DP, estimated with midlife exposure data, was highly correlated with the consumption of green-coloured fruits and vegetables, vegetable oils, orange-coloured fruits and vegetables and soup. Furthermore, it was positively associated with the composite cognitive performance score assessed 13 years later, even after accounting for confounders such as sociodemographic factors, lifestyle characteristics and health status. More specifically, high CDP scores were related to better episodic memory, semantic fluency, working memory and executive functioning. The positive association between this DP and subsequent cognitive performance adds support to previous research reporting better cognitive status( Reference Jama, Launer and Witteman 13 Reference Ortega, Requejo and Andres 15 ), lower cognitive decline( Reference Wengreen, Munger and Corcoran 19 , Reference Hu, Bretsky and Crimmins 20 ) or lower probability( Reference Rinaldi, Polidori and Metastasio 16 Reference Wang, Shinto and Connor 18 ) or risk of dementia( Reference Engelhart, Geerlings and Ruitenberg 21 ) among participants with high β-carotene intake or biomarker status.

However, our findings cannot be directly compared with those of prior research investigating associations between a wide variety of carotenoids and cognitive outcomes. For example, in the EVA (Etude du Vieillissement Artériel) study, some carotenoids, i.e. lycopene and zeaxanthin but not β-carotene, were found to be associated with specific cognitive tests mostly reflecting executive functioning. In the present study, the extracted CDP was not materially correlated with either lycopene or zeaxanthin, thus not allowing to draw a conclusion specifically with regard to these carotenoids. Other studies reporting results with regard to carotenoids did not find any association with cognitive function( Reference Schmidt, Hayn and Reinhart 22 , Reference Kang and Grodstein 24 ) or decline( Reference Kang and Grodstein 24 ). Besides, lower carotenoid status was observed among Alzheimer's disease cases compared with controls( Reference Wang, Shinto and Connor 18 , Reference Polidori, Mattioli and Aldred 34 ), but the case–control design of these studies prevented an inference about temporality or causality.

Lutein and zeaxanthin (xanthophylls, i.e. oxygenated carotenoids), which act as antioxidant and anti-inflammatory molecules, are preferentially accumulated in the macula of the retina and may be good candidates for eye/vision protection through various pathways( Reference Sabour-Pickett, Nolan and Loughman 51 ). They have also been posited to represent approximately 70 % of the carotenoids in the brain, and growing evidence favours a role of xanthophylls in the maintenance of cognitive function( Reference Johnson 33 ). In the present study, lutein was found to be strongly associated with CDP, whereas zeaxanthin exhibited a weaker association; thus, the present study does not allow drawing conclusions regarding the role of xanthophylls. Carefully designed studies are needed to test whether serum xanthophylls, which have been found to be correlated with xanthophylls present in the brain( Reference Johnson 33 ), exhibit neuroprotective effects.

Generally, traditional reductive approaches used in nutritional epidemiology may not capture the synergy between foods and nutrients( Reference Hoffmann, Schulze and Schienkiewitz 35 ). The RRR method yields a linear combination of food groups/nutrients maximising the proportion of the explained variation in intermediate response variables potentially implicated in the associations of interest. As such, the RRR method allows testing hypotheses regarding pathways of interest related to diet and diseases( Reference Schulze, Hoffmann and Manson 52 , Reference Schulze and Hoffmann 53 ). The RRR factor score is a combination of various food intakes influencing biomarker concentrations, but does not necessarily represent a combination of foods and drinks often consumed together. As with other dimension-reducing methods, the determination of the number of retained factors is often arbitrary. In the present study, we focused on the first extracted pattern that optimally explained the variation in carotenoid status. Indeed, the first pattern was positively correlated (>0·20) with most of the carotenoids and accounted for 11·66 % of the total variation, while other factors were much more weakly correlated with these biomarkers. In contrast, in the original study carried out by Hoffmann et al. ( Reference Hoffmann, Schulze and Schienkiewitz 35 ), food intakes explained more than 90 % of the variation in nutrient intakes. This is not surprising, since nutrient intakes are directly estimated from food intakes. Besides, when biomarkers are introduced as response variables, as in the present study, the percentage of variation in response is much weaker as reported previously( Reference McNaughton, Mishra and Brunner 54 Reference Heidemann, Hoffmann and Spranger 56 ).

A major limitation of the present study was the unavailability of baseline cognitive performance measurements, preventing an inference of causality. No baseline differences in cognitive performance according to DP can be assumed. However, cognitive impairment at baseline leading to modified DP (and thus leading to reverse causality) is unlikely, considering the relatively young age of the study population and the ability to follow the comprehensive study protocol (filling out many questionnaires). In addition, we extrapolated the CDP identified among a subsample to the entire sample. This may generate concerns about the generalisability of the DP and CDP data. However, we showed that the DP was relatively robust and that the subsample from which the DP was extracted was similar to the whole sample across many baseline and follow-up characteristics. Furthermore, some authors have suggested that the generalisability of RRR-extracted DP is possible as long as the populations of interest share similar profiles, in particular, with regard to dietary behaviour( Reference Imamura, Lichtenstein and Dallal 57 ). Next, caution is needed when generalising the present findings, as the participants were relatively healthy volunteers involved in a long-term nutritional study( Reference Rothman, Greenland and Lash 58 ). Another issue pertains to the additional potential bias due to non-response regarding the dietary questionnaires and/or cognitive evaluation. However, the use of inverse probability weighting to correct for potential selection bias did not substantially modify the main findings. Finally, residual confounding cannot be excluded regardless of the extensive adjustment for confounders.

The present study also exhibits strengths and other original aspects including its large sample of community-dwelling subjects, its long follow-up period and the use of highly accurate dietary data reflecting midlife exposure. The availability of biomarker data for a subsample of the population allowed the calculation of DP. Next, DP extracted using the RRR procedure allowed analysing associations between food group consumption and cognitive function beyond the existing associations between food consumption practices, as would be estimated by principal component analysis. Indeed, the RRR method is original and versatile. The use of biomarkers as response variables in the RRR procedure allowed exploring possible mechanistic pathways through which food group consumption and even specific DP could act on a health outcome. Thus, the present study revealed the food groups whose consumption directly contributed to plasma carotenoid status. Finally, our findings are of major interest from a public health viewpoint, since the use of the RRR procedure allows formulating easily understandable diet-based public health messages. In the cognitive domain, such messages are of utmost importance, given that prevention is a cost-effective strategy and the prevention of dementia should be initiated in middle age when potential cognitive disorders are presymptomatic( Reference de la Torre 1 , Reference Mortimer, Borenstein and Gosche 59 ).

In conclusion, the present study adds new support with regard to the positive association between a CDP in midlife and subsequent cognitive function, especially in terms of executive functioning and episodic memory, which are cognitive domains particularly vulnerable during the pathological ageing of the brain. Upon confirmation in other settings, these findings may argue that sufficient quantity and variety of coloured fruits and vegetables in one's diet may help to maintain the health of the brain during ageing.

Supplementary material

To view supplementary material for this article, please visit http://dx.doi.org/10.1017/S0007114513003188

Acknowledgements

The authors cordially thank Stéphane Raffard, neuropsychologist, who was responsible for standardisation of the cognitive evaluation; Nathalie Arnault (statistician), who coordinated data management; Frédérique Ferrat, who coordinated the logistic aspects of the neuropsychological evaluation; and Gwenaël Monot (computer scientist), who coordinated the computing aspects. The present study was supported by the ANR (National Research Agency; grant no. ANR-05-PNRA-010) and DGS (Ministry of Health). The ANR and DGS had no role in the design and analysis of the study or in the writing of this article. The authors' contributions were as follows: E. K.-G. carried out data checking and analyses and was responsible for manuscript drafting; V. A. A., C. J., C. J., S. H. and P. G. were involved in the interpretation of the results and editing of the manuscript; E. K.-G., P. G. and S. H. were responsible for developing the design and protocol of the study. All authors read and approved the final version of the manuscript. None of the authors has any conflicts of interest.

References

1 de la Torre, JC (2010) Alzheimer's disease is incurable but preventable. J Alzheimers Dis 20, 861870.Google Scholar
2 Middleton, LE & Yaffe, K (2009) Promising strategies for the prevention of dementia. Arch Neurol 66, 12101215.Google Scholar
3 Daviglus, ML, Plassman, BL, Pirzada, A, et al. (2011) Risk factors and preventive interventions for Alzheimer disease: state of the science. Arch Neurol 68, 11851190.Google Scholar
4 Coley, N, Andrieu, S, Gardette, V, et al. (2008) Dementia prevention: methodological explanations for inconsistent results. Epidemiol Rev 30, 3566.Google Scholar
5 Plassman, BL, Williams, JW Jr, Burke, JR, et al. (2010) Systematic review: factors associated with risk for and possible prevention of cognitive decline in later life. Ann Intern Med 153, 182193.CrossRefGoogle ScholarPubMed
6 Gemma, C, Vila, J, Bachstetter, A, et al. (2007) Oxidative stress and the aging brain: from theory to prevention. In Brain Aging: Models, Methods, and Mechanisms [Riddle, DR, editor]. Boca Raton, FL: CRC Press.Google Scholar
7 Polidori, MC, De Spirt, S, Stahl, W, et al. (2012) Conflict of evidence: carotenoids and other micronutrients in the prevention and treatment of cognitive impairment. Biofactors 38, 167171.Google Scholar
8 Federico, A, Cardaioli, E, Da, PP, et al. (2012) Mitochondria, oxidative stress and neurodegeneration. J Neurol Sci 322, 254262.Google Scholar
9 Radak, Z, Zhao, Z, Goto, S, et al. (2011) Age-associated neurodegeneration and oxidative damage to lipids, proteins and DNA. Mol Aspects Med 32, 305315.Google Scholar
10 Patten, DA, Germain, M, Kelly, MA, et al. (2010) Reactive oxygen species: stuck in the middle of neurodegeneration. J Alzheimers Dis 20, Suppl. 2, S357S367.Google Scholar
11 Donaldson, MS (2011) A carotenoid health index based on plasma carotenoids and health outcomes. Nutrients 3, 10031022.Google Scholar
12 Giordano, P, Scicchitano, P, Locorotondo, M, et al. (2012) Carotenoids and cardiovascular risk. Curr Pharm Des 18, 55775589.Google Scholar
13 Jama, JW, Launer, LJ, Witteman, JC, et al. (1996) Dietary antioxidants and cognitive function in a population-based sample of older persons. The Rotterdam Study. Am J Epidemiol 144, 275280.Google Scholar
14 Akbaraly, NT, Faure, H, Gourlet, V, et al. (2007) Plasma carotenoid levels and cognitive performance in an elderly population: results of the EVA Study. J Gerontol A Biol Sci Med Sci 62, 308316.Google Scholar
15 Ortega, RM, Requejo, AM, Andres, P, et al. (1997) Dietary intake and cognitive function in a group of elderly people. Am J Clin Nutr 66, 803809.Google Scholar
16 Rinaldi, P, Polidori, MC, Metastasio, A, et al. (2003) Plasma antioxidants are similarly depleted in mild cognitive impairment and in Alzheimer's disease. Neurobiol Aging 24, 915919.Google Scholar
17 von Arnim, CA, Herbolsheimer, F, Nikolaus, T, et al. (2012) Dietary antioxidants and dementia in a population-based case–control study among older people in south Germany. J Alzheimers Dis 31, 717724.Google Scholar
18 Wang, W, Shinto, L, Connor, WE, et al. (2008) Nutritional biomarkers in Alzheimer's disease: the association between carotenoids, n-3 fatty acids, and dementia severity. J Alzheimers Dis 13, 3138.Google Scholar
19 Wengreen, HJ, Munger, RG, Corcoran, CD, et al. (2007) Antioxidant intake and cognitive function of elderly men and women: the Cache County Study. J Nutr Health Aging 11, 230237.Google Scholar
20 Hu, P, Bretsky, P, Crimmins, EM, et al. (2006) Association between serum beta-carotene levels and decline of cognitive function in high-functioning older persons with or without apolipoprotein E 4 alleles: MacArthur studies of successful aging. J Gerontol A Biol Sci Med Sci 61, 616620.Google Scholar
21 Engelhart, MJ, Geerlings, MI, Ruitenberg, A, et al. (2002) Dietary intake of antioxidants and risk of Alzheimer disease. JAMA 287, 32233229.Google Scholar
22 Schmidt, R, Hayn, M, Reinhart, B, et al. (1998) Plasma antioxidants and cognitive performance in middle-aged and older adults: results of the Austrian Stroke Prevention Study. J Am Geriatr Soc 46, 14071410.Google Scholar
23 Kalmijn, S, Feskens, EJ, Launer, LJ, et al. (1997) Polyunsaturated fatty acids, antioxidants, and cognitive function in very old men. Am J Epidemiol 145, 3341.Google Scholar
24 Kang, JH & Grodstein, F (2008) Plasma carotenoids and tocopherols and cognitive function: a prospective study. Neurobiol Aging 29, 13941403.Google Scholar
25 McNeill, G, Jia, X, Whalley, LJ, et al. (2011) Antioxidant and B vitamin intake in relation to cognitive function in later life in the Lothian Birth Cohort 1936. Eur J Clin Nutr 65, 619626.Google Scholar
26 Laurin, D, Masaki, KH, Foley, DJ, et al. (2004) Midlife dietary intake of antioxidants and risk of late-life incident dementia: the Honolulu-Asia Aging Study. Am J Epidemiol 159, 959967.Google Scholar
27 Morris, MC, Evans, DA, Bienias, JL, et al. (2002) Dietary intake of antioxidant nutrients and the risk of incident Alzheimer disease in a biracial community study. JAMA 287, 32303237.Google Scholar
28 Devore, EE, Grodstein, F, van Rooij, FJ, et al. (2010) Dietary antioxidants and long-term risk of dementia. Arch Neurol 67, 819825.Google Scholar
29 Kang, JH, Cook, NR, Manson, JE, et al. (2009) Vitamin E, vitamin C, beta carotene, and cognitive function among women with or at risk of cardiovascular disease: The Women's Antioxidant and Cardiovascular Study. Circulation 119, 27722780.Google Scholar
30 Wolters, M, Hermann, S & Hahn, A (2004) Effects of 6-month multivitamin supplementation on serum concentrations of alpha-tocopherol, beta-carotene, and vitamin C in healthy elderly women. Int J Vitam Nutr Res 74, 161168.Google Scholar
31 Yaffe, K, Clemons, TE, McBee, WL, et al. (2004) Impact of antioxidants, zinc, and copper on cognition in the elderly: a randomized, controlled trial. Neurology 63, 17051707.Google Scholar
32 Grodstein, F, Kang, JH, Glynn, RJ, et al. (2007) A randomized trial of beta carotene supplementation and cognitive function in men: the Physicians' Health Study II. Arch Intern Med 167, 21842190.Google Scholar
33 Johnson, EJ (2012) A possible role for lutein and zeaxanthin in cognitive function in the elderly. Am J Clin Nutr 96, 1161S1165S.Google Scholar
34 Polidori, MC, Mattioli, P, Aldred, S, et al. (2004) Plasma antioxidant status, immunoglobulin g oxidation and lipid peroxidation in demented patients: relevance to Alzheimer disease and vascular dementia. Dement Geriatr Cogn Disord 18, 265270.Google Scholar
35 Hoffmann, K, Schulze, MB, Schienkiewitz, A, et al. (2004) Application of a new statistical method to derive dietary patterns in nutritional epidemiology. Am J Epidemiol 159, 935944.Google Scholar
36 Hercberg, S, Galan, P, Preziosi, P, et al. (2004) The SU.VI.MAX Study: a randomized, placebo-controlled trial of the health effects of antioxidant vitamins and minerals. Arch Intern Med 164, 23352342.Google Scholar
37 Hercberg, S, Preziosi, P, Briancon, S, et al. (1998) A primary prevention trial using nutritional doses of antioxidant vitamins and minerals in cardiovascular diseases and cancers in a general population: the SU.VI.MAX study – design, methods, and participant characteristics. SUpplementation en VItamines et Mineraux AntioXydants. Control Clin Trials 19, 336351.Google Scholar
38 Steghens, JP, van Kappel, AL, Riboli, E, et al. (1997) Simultaneous measurement of seven carotenoids, retinol and alpha-tocopherol in serum by high-performance liquid chromatography. J Chromatogr B Biomed Sci Appl 694, 7181.Google Scholar
39 Le Moullec, N, Deheeger, M, Preziosi, P, et al. (1996) Validation du manuel photos utilisé pour l'enquête alimentaire de l'étude SU.VI.MAX. Cahier de Nutrition et de Diététique 31, 158164.Google Scholar
40 Hercberg, S (coordinator) (2005) Table de composition SU.VI.MAX des aliments (Table Food Composition SU.VI.MAX). Paris: Les éditions INSERM/Economica, 182 pp.Google Scholar
41 Ivanoiu, A, Adam, S, Van der Linden, LM, et al. (2005) Memory evaluation with a new cued recall test in patients with mild cognitive impairment and Alzheimer's disease. J Neurol 252, 4755.Google Scholar
42 Lezak, MD, Howieson, DB & Loring, DW (2004) Neuropsychological Assessment, 4th ed. New York, NY: Oxford University Press.Google Scholar
43 Wechsler, D (1981) Wechsler Adult Intelligence Scale-Revised. New York, NY: Psychological Corporation.Google Scholar
44 Delis, DC, Kaplan, E & Kramer, JH (2001) Delis–Kaplan Executive Function System (D-KEFS) Examiner's Manual. San Antonio, TX: The Psychological Corporation.Google Scholar
45 Kesse-Guyot, E, Andreeva, VA, Jeandel, C, et al. (2012) A healthy dietary pattern at midlife is associated with subsequent cognitive performance. J Nutr 142, 909915.Google Scholar
46 Radloff, L (1977) The CES-D scale: a self-report depression scale for research in the general population. Appl Psychol Meas 1, 385401.Google Scholar
47 Willett, WC (1998) Nutritional Epidemiology, 2nd ed. New York, NY: Oxford University Press.Google Scholar
48 Cole, SR & Hernan, MA (2008) Constructing inverse probability weights for marginal structural models. Am J Epidemiol 168, 656664.Google Scholar
49 Seaman, SR & White, IR (2011) Review of inverse probability weighting for dealing with missing data. Stat Methods Med Res 22, 278295.Google Scholar
50 Schulze, MB, Hoffmann, K, Kroke, A, et al. (2003) An approach to construct simplified measures of dietary patterns from exploratory factor analysis. Br J Nutr 89, 409419.Google Scholar
51 Sabour-Pickett, S, Nolan, JM, Loughman, J, et al. (2012) A review of the evidence germane to the putative protective role of the macular carotenoids for age-related macular degeneration. Mol Nutr Food Res 56, 270286.Google Scholar
52 Schulze, MB, Hoffmann, K, Manson, JE, et al. (2005) Dietary pattern, inflammation, and incidence of type 2 diabetes in women. Am J Clin Nutr 82, 675684.Google Scholar
53 Schulze, MB & Hoffmann, K (2006) Methodological approaches to study dietary patterns in relation to risk of coronary heart disease and stroke. Br J Nutr 95, 860869.Google Scholar
54 McNaughton, SA, Mishra, GD & Brunner, EJ (2008) Dietary patterns, insulin resistance, and incidence of type 2 diabetes in the Whitehall II Study. Diabetes Care 31, 13431348.Google Scholar
55 Hoffmann, K, Zyriax, BC, Boeing, H, et al. (2004) A dietary pattern derived to explain biomarker variation is strongly associated with the risk of coronary artery disease. Am J Clin Nutr 80, 633640.Google Scholar
56 Heidemann, C, Hoffmann, K, Spranger, J, et al. (2005) A dietary pattern protective against type 2 diabetes in the European Prospective Investigation into Cancer and Nutrition (EPIC) – Potsdam Study cohort. Diabetologia 48, 11261134.Google Scholar
57 Imamura, F, Lichtenstein, AH, Dallal, GE, et al. (2009) Generalizability of dietary patterns associated with incidence of type 2 diabetes mellitus. Am J Clin Nutr 90, 10751083.Google Scholar
58 Rothman, KJ, Greenland, S & Lash, TL (2008) Modern Epidemiology, 2nd ed. Philadelphia, PA: Lippincott Williams & Wilkins.Google Scholar
59 Mortimer, JA, Borenstein, AR, Gosche, KM, et al. (2005) Very early detection of Alzheimer neuropathology and the role of brain reserve in modifying its clinical expression. J Geriatr Psychiatry Neurol 18, 218223.Google Scholar
Figure 0

Table 1 Explained variation in the consumption of foods and in plasma carotenoid concentrations with the carotenoid-rich dietary pattern (n 381)

Figure 1

Table 2 Food groups associated with carotenoid-rich dietary patterns (extracted from n 2983)

Figure 2

Table 3 Baseline characteristics of the population across the tertiles of the reduced rank regression-extracted carotenoid-rich dietary pattern score (n 2983)* (Mean values and standard deviations)

Figure 3

Table 4 Associations between the carotenoid-rich dietary pattern score (in tertiles (T) and as continuous variable) and cognitive performance (Mean differences and 95 % confidence intervals)

Supplementary material: File

Kesse-Guyot Supplementary Material

Table S1

Download Kesse-Guyot Supplementary Material(File)
File 136.2 KB