Abstract
Fish are a primary source of long-chain omega-3 fatty acids, which may help delay cognitive aging. We pooled participants from the French Three-City study and 4 US cohorts (Nurses’ Health Study, Women’s Health Study, Chicago Health and Aging Project, and Rush Memory and Aging Project) for whom diet and cognitive data were available (n = 23,688 white persons, aged ≥65 years, 88% female, baseline year range of 1992–1999, and median follow-up range of 3.9–9.1 years) to investigate the relationship of fish intake to cognitive decline and examine interactions with genes related to Alzheimer disease. We estimated cohort-specific associations between fish and change in composite scores of global cognition and episodic memory using linear mixed models, and we pooled results using inverse-variance weighted meta-analysis. In multivariate analyses, higher fish intake was associated with slower decline in both global cognition and memory (P for trend ≤ 0.031). Consuming ≥4 servings/week versus <1 serving/week of fish was associated with a lower rate of memory decline: 0.018 (95% confidence interval: 0.004, 0.032) standard units, an effect estimate equivalent to that found for 4 years of age. For global cognition, no comparisons of higher versus low fish intake reached statistical significance. In this meta-analysis, higher fish intake was associated with a lower rate of memory decline. We found no evidence of effect modification by genes associated with Alzheimer disease.
Keywords: Alzheimer dementia, cognitive aging, gene-environment interaction, omega-3 fatty acids
Fish are a primary source of marine long-chain omega-3 polyunsaturated fatty acids (LCn-3 PUFAs), which are promising candidates in research on delaying cognitive aging and reducing the risk of Alzheimer disease (AD). Docosahexaenoic acid (DHA), a LCn-3 PUFA, is a requirement for the development of the mammalian brain—in particular for the hippocampus, a key structure for episodic memory, which is altered in both aging and AD. Furthermore, human autopsy studies have reported 30%–50% lower DHA levels in the hippocampus of AD patients compared with controls (1–3). Hence, an adequate supply of essential LCn-3 PUFAs in the diet (mainly from fish) may be critical for the maintenance of cognition and memory with aging. Moreover, fish contains vitamin D and selenium—2 potentially neuroprotective nutrients that may add to the benefit of LCn-3 PUFAs. Given the burden of cognitive aging and AD worldwide, delaying cognitive decline by only a few years with fairly simple approaches, such as increasing fish intake, may have a large public health impact (4).
Some observational studies have found that intakes of fish, or higher concentrations of LCn-3 PUFA in blood, may be related to a lower risk of dementia and to slower cognitive decline (5–9). However, inconsistent findings have been reported, especially with cognitive decline (10), and randomized trials have been inconclusive to date (11). For example, 2 small trials in healthy subjects have demonstrated the efficacy of approximately 1 g LCn-3PUFA per day in preserving episodic memory but in no other cognitive domain (12, 13). However, episodic memory has not been well-explored in observational research. It is also possible that current research inadequately targets individuals according to their genetic susceptibility to AD. Indeed, many AD genes (e.g., apolipoprotein E gene (APOE), clusterin gene (CLU), and complement receptor 1 gene (CR1)) are involved in lipid metabolism or inflammation, a likely pathway by which LCn-3PUFAs may act (14).
In this study, we pooled selected participants from 5 large cohorts of older subjects (the Three-City (3C) study, the Nurses’ Health Study (NHS), the Women’s Health Study (WHS), the Chicago Health and Aging Project (CHAP), and the Rush Memory and Aging Project (MAP)), for all of whom there were data on diet and repeated measures of cognition, to conduct a statistically powerful investigation of fish intake and decline in global cognition and episodic memory. We secondarily examined potential interactions of fish with APOE ε4 variant (APOEε4) status and 13 candidate single nucleotide polymorphisms (SNPs) associated with AD.
METHODS
Study populations
This analysis included participants from a large French cohort, the 3C study, and 4 large US cohorts: NHS, WHS, CHAP, and MAP. A detailed description of the cohorts has been provided in Web Appendix 1 (available at https://academic.oup.com/aje). Briefly, the 3C study was initiated in 1999–2000 among 9,294 older persons, aged ≥65 years and not institutionalized, in 3 cities in France (Bordeaux, Dijon, and Montpellier); the initial purpose of the 3C study was to study vascular risk factors for dementia. The NHS started in 1976 among 121,700 US female nurses aged 30–55 years; a cognitive substudy was initiated between 1995 and 2001 among nurses aged ≥70 years and free of stroke. The WHS was a randomized controlled trial of low-dose aspirin and vitamin E supplements for the primary prevention of cardiovascular disease and cancer, initiated between 1992 and 1995 among 39,876 US female health professionals aged ≥45 years; a cognitive substudy was conducted an average 5.6 years after trial baseline among 6,377 WHS participants aged ≥65 years. CHAP and MAP are 2 US cohort studies of older persons, conducted in Chicago. CHAP included participants aged ≥65 years in 3 south-side neighborhoods of Chicago, and MAP included volunteers aged >55 years, free of known dementia and living in retirement communities, senior public housing, or individual homes.
In the present study, we included reported white participants across the 5 cohorts, who: 1) completed at least 1 food frequency questionnaire (FFQ) before initial cognitive assessment; 2) had at least 1 repeated cognitive evaluation over 4–9 years; 3) had available data for APOE genotype and other candidate SNPs; and 4) had no missing information for fish intake or educational level. In the NHS, where genome-wide genotyping was performed in <25% of those included in the cognitive substudy, to retain study power, primary analyses of fish and cognitive function were conducted on the total sample rather than on only the genotyped sample. Details on the sample selection in each cohort have been provided in Web Appendix 1. Overall, there was no relationship between diet, cognition, and selection into the genotyped subsets within any of the cohorts, and thus meaningful selection bias in these analyses is unlikely.
Assessment of cognitive function
Cognitive testing was performed by trained, in-person interviewers in the 3C study, CHAP, and MAP and in validated telephone interviews in the NHS and WHS. Although the cognitive tests differed across cohorts, the overall batteries reflected similar cognitive constructs. In particular, the telephone instrument used in the NHS and WHS was designed to be similar to that used in the 2 Rush cohorts (CHAP and MAP); in a validation study in which a group of participants was administered both the Rush battery in person and the NHS battery by phone, a high correlation (ρ > 0.80) was found between overall performance (averaging all component tests), demonstrating high comparability of the 2 assessment batteries.
We defined composite scores of global cognition and episodic memory as primary outcomes. A global cognitive score was computed in each cohort as the mean of z scores of cognitive tests assessing: 1) global cognition, 2) verbal fluency/semantic memory, and 3) working memory and attention. In addition, episodic memory tests were combined in a composite score, calculated as the mean of z scores of tests assessing episodic memory. The cognitive tests used in each cohort are detailed in Web Appendix 2.
Assessment of diet
In these analyses, when FFQs were administered repeatedly during cohort follow-up (3C study, NHS, CHAP, and MAP), we used the FFQ closest to the initial cognitive evaluation for ascertainment of diet. Assessment of habitual consumption of foods and beverages was largely similar across the 4 US cohorts. In these FFQs, frequency of consumption was queried, with possible responses ranging from “never or <1 time/month” to “≥6 times/day.” In the 3C study, habitual intake of 10 food categories was assessed using a brief qualitative FFQ. Frequency of consumption was recorded in 6 categories: never, <1/week, 1/week, 2–3/week, 4–6/week, and ≥6/week. Dietary intakes obtained with this brief questionnaire showed reasonable correlations with intakes estimated with a comprehensive FFQ administered in a subsample in Bordeaux (e.g., ρ = 0.52 for fish intake), and we previously found significant associations between dietary habits assessed with this questionnaire and the risk of dementia (15).
In the 4 US cohorts, weekly consumption of fish was computed by summing the responses to the fish items in the FFQ (in the NHS and the WHS, canned tuna fish, dark meat fish, and other fish; in CHAP and MAP, tuna fish sandwich, fish sticks/fish cakes/fish sandwich, and fresh fish as a main dish). To harmonize data for fish intake across cohorts, we converted weekly servings of fish intake in US cohorts and the 6 categories of intake in the 3C study into a similar 4-category variable of servings per week: <1, 1, 2–3, and ≥4.
The coding for other dietary variables used in multivariate models (meat, fruits and vegetables, and energy intakes) is detailed in Web Appendix 3.
Assessment of genotype data
APOEε4 status was defined as carrying at least one ε4 allele for the APOE gene versus no ε4 allele. Furthermore, genetic information included genotyped and imputed SNPs obtained from previous genome-wide association studies in each cohort. We focused on 13 SNPs at 11 loci, selected as those most consistently reported in AD genome-wide association studies (listed in Web Table 1; minor allele frequencies provided in Web Table 2) (16–18).
Other variables
Covariates were obtained from the questionnaire closest to dietary assessment. These included: 1) primary potential confounders (age, sex (when applicable), education, income, regular exercise, smoking, and moderate alcohol intake); and 2) variables that might be confounders and/or mediators (body mass index; history of depression or depressive symptomatology (ascertained using the Center for Epidemiologic Studies-Depression (CES-D) scale (19)); and history of heart disease, hypertension, hypercholesterolemia, or type 2 diabetes (either self-reported, as in NHS, WHS, CHAP, MAP, or based on direct measures, as in the 3C study and MAP)).
Statistical analyses
In primary analyses, in each cohort, we modeled the trajectories of repeated cognitive scores across fish intake categories using linear mixed models (20). The models included an intercept that represented the cognitive level at baseline and a slope that represented the annual change in scores over time as well as a random intercept and random slope to account for interindividual variability. We examined linear trends across fish intake categories using a discrete variable taking 4 possible values (0, 1, 2.5, or 4 servings/week). We pooled the mean differences in slopes of cognitive change for: 1) each fish intake category (1, 2–3, and ≥4 times/week) compared with the referent (<1 time/week); and 2) each additional discrete weekly fish serving, using inverse-variance weighted, fixed-effects meta-analyses.
Other analyses investigated interactions between fish intake and APOEε4/candidate SNPs on the slopes of change in global cognition and episodic memory using additive genetic models (i.e., regressions of phenotype on the number of reference alleles (0, 1, or 2)) or, equivalently, the imputed dosage for imputed genotypes. Interaction terms for APOEε4 and each SNP on both the intercept and the slope of cognitive change were constructed by creating cross-product terms of: 1) fish intake (discrete, 0, 1, 2.5, or 4 servings/week) × the candidate SNP (discrete, 0, 1, or 2 minor alleles; or continuous, imputation dosage), and 2) fish intake × candidate SNP × time. Both interaction terms were added to separate linear mixed models (one for each SNP), adjusted for age, sex (when applicable), education, study center (3C study), and treatment allocation arm (WHS). As in the main analysis, in secondary analyses the findings were pooled using inverse-variance weighted meta-analyses.
RESULTS
Primary analyses on fish and cognitive change included 23,688 participants across the 5 cohorts (Table 1). Mean age ranged from 71.9 years in the WHS to 82.2 years in MAP; 2 cohorts (NHS and WHS) included only women. Weekly consumption of fish varied across cohorts, with higher intakes in the 3C study and MAP compared with NHS, WHS, and CHAP. For example, about 50% of the 3C study and MAP samples reported eating fish ≥2 times a week, compared with 38% or less in the 3 other cohorts (Table 1). Accordingly, more participants consumed fish less than once a week in NHS, WHS, and CHAP (65%, 50% and 35%, respectively) than in the 3C study and MAP (11% and 25%, respectively).
Table 1.
Characteristics of Participants From 5 Cohort Studies Included in a Meta-Analysis of the Association of Fish Intake and Alzheimer Disease Genes With Cognitive Decline (n = 23,688), France and the United States, 1992–2012
| Cohort | 3C Study (n = 5,641), % | NHS (n = 13,129), % | WHS (n = 3,170), % | CHAP (n = 932), % | MAP (n = 816), % |
|---|---|---|---|---|---|
| Follow-up for cognition after ascertainment of diet, yearsa | 6.7 (3.9, 7.1) | 6.2 (4.5, 6.5) | 3.9 (3.7, 4.0) | 9.1 (3.5, 15.3) | 5.0 (3.0, 7.0) |
| Age at dietary assessment, yearsb | 73.9 (5.3) | 74.2 (2.4) | 71.9 (4.1) | 73.4 (6.0) | 82.2 (6.9) |
| Female sex | 61 | 100 | 100 | 61 | 73 |
| Fish intake categories, servings/week | |||||
| <1 | 11 | 65 | 50 | 35 | 25 |
| 1 | 38 | 24 | 31 | 27 | 25 |
| 2–3 | 45 | 9 | 12 | 28 | 43 |
| ≥4 | 6 | 2 | 7 | 10 | 6 |
Abbreviations: 3C study, Three-City study; CHAP, Chicago Health and Aging Project; MAP, Memory and Aging Project; NHS, Nurses’ Health Study; WHS, Women’s Health Study.
a Values are expressed as median (interquartile range).
b Values are expressed as mean (standard deviation).
Fish intake and decline in global cognition and episodic memory
Median follow-up for cognition after ascertainment of diet ranged from 3.9 years (in WHS) to 9.1 years (in CHAP; Table 1). Average cognitive performances declined over time in all cohorts (for global cognition, slopes ranged from −0.17 (95% confidence interval (CI): −0.19, −0.15) standard units/year in MAP to −0.02 (95% CI: −0.03, −0.01) standard units/year in WHS in unadjusted linear mixed models; for episodic memory, slopes ranged from −0.10 (95% CI: −0.11, −0.08) standard units/year in MAP to −0.01 (95% CI: −0.03, 0.00) standard units/year in WHS). When we investigated the relationship of fish intake to cognition, we did not find any significant association between fish intake and baseline cognitive level (P for trend = 0.515 for global cognition and 0.057 for episodic memory, in pooled analyses adjusted for confounders). When we examined rate of change in the global cognitive score (Table 2), we did not find significant relationships between fish intake and cognitive decline within any of the individual cohorts. However, when we combined the cohorts, we found a significant trend such that increasing intake of fish was associated with slower global cognitive decline (pooled P for trend = 0.031 in multivariate analysis). In general, effect estimates for specific categories of intake were modest.
Table 2.
Multivariate Associations Between Fish Intake and Change in Global Cognition Across 5 Cohort Studies, France and the United States, 1992–2012
| Cohort and Weekly Servings of Fish | No. of Participants | Global Cognitive Change (Standard Units/year) | |||||
|---|---|---|---|---|---|---|---|
| Model 1a | Model 2b | ||||||
| β | 95% CI | P for Trendc | β | 95% CI | P for Trendc | ||
| 3C study | 5,641 | 0.225 | 0.380 | ||||
| <1 | 629 | 0 | Referent | 0 | Referent | ||
| 1 | 2,178 | −0.001 | −0.012, 0.009 | −0.002 | −0.013, 0.009 | ||
| 2–3 | 2,519 | 0.002 | −0.009, 0.012 | 0.000 | −0.010, 0.011 | ||
| ≥4 | 315 | 0.008 | −0.008, 0.024 | 0.006 | −0.010, 0.023 | ||
| NHS | 13,129 | 0.070 | 0.130 | ||||
| <1 | 8,502 | 0 | Referent | 0 | Referent | ||
| 1 | 3,099 | 0.003 | −0.002, 0.008 | 0.002 | −0.003, 0.008 | ||
| 2–3 | 1,184 | 0.008 | −0.000, 0.015 | 0.007 | −0.001, 0.015 | ||
| ≥4 | 344 | 0.003 | −0.011, 0.016 | 0.003 | −0.011, 0.017 | ||
| WHS | 3,170 | 0.608 | 0.675 | ||||
| <1 | 1,586 | 0 | Referent | 0 | Referent | ||
| 1 | 987 | −0.017 | −0.033, −0.000 | −0.017 | −0.034, −0.000 | ||
| 2–3 | 367 | −0.012 | −0.036, 0.012 | −0.014 | −0.038, 0.011 | ||
| ≥4 | 230 | 0.022 | −0.007, 0.051 | 0.020 | −0.009, 0.050 | ||
| CHAP | 932 | 0.156 | 0.077 | ||||
| <1 | 327 | 0 | Referent | 0 | Referent | ||
| 1 | 250 | 0.002 | −0.019, 0.022 | 0.004 | −0.016, 0.024 | ||
| 2–3 | 338 | 0.012 | −0.007, 0.032 | 0.017 | −0.003, 0.037 | ||
| ≥4 | 17 | 0.022 | −0.047, 0.091 | 0.019 | −0.051, 0.089 | ||
| MAP | 816 | 0.869 | 0.732 | ||||
| <1 | 204 | 0 | Referent | 0 | Referent | ||
| 1 | 206 | 0.043 | −0.009, 0.095 | 0.053 | 0.001, 0.105 | ||
| 2–3 | 354 | 0.029 | −0.018, 0.076 | 0.032 | −0.018, 0.082 | ||
| ≥4 | 52 | −0.019 | −0.104, 0.067 | 0.002 | −0.086, 0.091 | ||
| Pooled | 23,688 | 0.009 | 0.031 | ||||
| <1 | 11,248 | 0 | Referent | 0 | Referent | ||
| 1 | 6,720 | −0.001 | −0.010, 0.008 | −0.001 | −0.009, 0.009 | ||
| 2–3 | 4,762 | 0.005 | −0.001, 0.011 | 0.005 | −0.003, 0.013 | ||
| ≥4 | 958 | 0.007 | −0.003, 0.017 | 0.006 | −0.004, 0.016 | ||
Abbreviations: 3C study, Three-City study; CHAP, Chicago Health and Aging Project; CI, confidence interval; MAP, Memory and Aging Project; NHS, Nurses’ Health Study; WHS, Women’s Health Study.
a Model 1 adjusted for age, education, sex (in 3C study, CHAP, and MAP only, where applicable), study center (3C study only), and treatment allocation arm (WHS only).
b Model 2 included covariates from model 1 and income, smoking, regular exercise, total energy intake, intakes of fruits and vegetables and meat, and moderate alcohol consumption.
c Estimates were computed in each cohort using linear mixed models and further pooled using inverse-variance weighted meta-analysis. P-for-trend values were computed using fish intake as a discrete variable taking 4 possible values (servings/week: 0, 1, 2.5, or 4).
When we examined the rate of decline in episodic memory (Table 3), fish intake was generally not significantly associated with slower decline after adjustment in individual cohorts; however, as with global cognition, when we pooled the cohorts we found significant trends of increasing fish intake associated with decreasing rate of episodic memory decline (pooled P for trend = 0.024). Moreover, we found a statistically significant relationship of fish intake to lower rate of decline for those consuming ≥4 servings of fish per week vs <1 serving per week (mean difference = 0.018 (95% CI: 0.004, 0.032) standard units). To help interpret these associations, we contrasted the effect estimate we found for age across the cohorts, with the effect estimate for fish intake and episodic memory decline in pooled analyses. Specifically, ≥4 servings of fish per week versus <1 serving/week was cognitively equivalent to 4 years of aging; that is, older persons consuming at least 4 weekly servings of fish had memory decline similar to those who were 4 years younger in age.
Table 3.
Multivariate Associations Between Fish Intake and Change in Episodic Memory Across 5 Cohort Studies, France and the United States, 1992–2012
| Cohort and Weekly Servings of Fish | No. of Participants | Episodic Memory Change (Standard Units/year) | |||||
|---|---|---|---|---|---|---|---|
| Model 1a | Model 2b | ||||||
| β | 95% CI | P for Trendc | β | 95% CI | P for Trendc | ||
| 3C study | 5,641 | 0.037 | 0.116 | ||||
| <1 | 629 | 0 | Referent | 0 | Referent | ||
| 1 | 2,178 | 0.004 | −0.013, 0.022 | 0.001 | −0.017, 0.019 | ||
| 2–3 | 2,519 | 0.011 | −0.007, 0.028 | 0.006 | −0.011, 0.024 | ||
| ≥4 | 315 | 0.026 | −0.000, 0.053 | 0.022 | −0.006, 0.049 | ||
| NHS | 13,129 | 0.076 | 0.162 | ||||
| <1 | 8,502 | 0 | Referent | 0 | Referent | ||
| 1 | 3,099 | 0.001 | −0.005, 0.007 | −0.000 | −0.006, 0.006 | ||
| 2–3 | 1,184 | 0.006 | −0.003, 0.014 | 0.005 | −0.004, 0.013 | ||
| ≥4 | 344 | 0.010 | −0.005, 0.026 | 0.010 | −0.006, 0.025 | ||
| WHS | 3,170 | 0.002 | 0.007 | ||||
| <1 | 1,586 | 0 | Referent | 0 | Referent | ||
| 1 | 987 | −0.000 | −0.016, 0.016 | −0.001 | −0.018, 0.015 | ||
| 2–3 | 367 | 0.017 | −0.007, 0.040 | 0.015 | −0.009, 0.039 | ||
| ≥4 | 230 | 0.045 | 0.017, 0.073 | 0.041 | 0.012, 0.070 | ||
| CHAP | 932 | 0.853 | 0.771 | ||||
| <1 | 327 | 0 | Referent | 0 | Referent | ||
| 1 | 250 | −0.004 | −0.023, 0.016 | −0.001 | −0.020, 0.019 | ||
| 2–3 | 338 | −0.001 | −0.019, 0.018 | 0.001 | −0.018, 0.020 | ||
| ≥4 | 17 | 0.021 | −0.049, 0.090 | 0.022 | −0.049, 0.093 | ||
| MAP | 816 | 0.948 | 0.898 | ||||
| <1 | 204 | 0 | Referent | 0 | Referent | ||
| 1 | 206 | 0.018 | −0.011, 0.047 | 0.020 | −0.009, 0.049 | ||
| 2–3 | 354 | 0.010 | −0.016, 0.036 | 0.007 | −0.020, 0.035 | ||
| ≥4 | 52 | −0.011 | −0.060, 0.038 | −0.005 | −0.056, 0.046 | ||
| Pooled | 23,688 | 0.023 | 0.024 | ||||
| <1 | 11,248 | 0 | Referent | 0 | Referent | ||
| 1 | 6,720 | 0.001 | −0.004, 0.006 | 0.001 | −0.005, 0.006 | ||
| 2–3 | 4,762 | 0.007 | 0.001, 0.013 | 0.006 | −0.001, 0.012 | ||
| ≥4 | 958 | 0.020 | 0.003, 0.037 | 0.018 | 0.004, 0.032 | ||
Abbreviations: 3C study, Three-City study; CHAP, Chicago Health and Aging Project; CI, confidence interval; MAP, Memory and Aging Project; NHS, Nurses’ Health Study; WHS, Women’s Health Study.
a Model 1 adjusted for age, education, sex (in 3C study, CHAP, and MAP only, where applicable), study center (3C study only), and treatment allocation arm (WHS only).
b Model 2 included covariates from model 1 and income, smoking, regular exercise, total energy intake, intakes of fruits and vegetables and meat, and moderate alcohol consumption.
c Estimates were computed in each cohort using linear mixed models and further pooled using inverse-variance weighted meta-analysis. P-for-trend values were computed using fish intake as a discrete variable taking 4 possible values (servings/week: 0, 1, 2.5, or 4).
Additionally, when we further included vascular conditions in models, estimates were virtually unchanged (model 3, Web Tables 3 and 4). Likewise, additional adjustment for body mass index and depression did not influence findings (model 4, Web Tables 3 and 4).
Fish × gene interactions on decline in global cognition and episodic memory
We found no significant interaction between fish intake and APOEε4 or the SNPs on cognitive change (Web Figure 1). Annual changes in both global cognition and episodic memory for each additional weekly fish serving per additional copy of a SNP allele did not significantly differ from zero for any of the genetic variants, either for global cognition (Web Figure 1, panel a) or for episodic memory (Web Figure 1, panel b) in multivariate models.
DISCUSSION
In this pooled analysis, including selected participants from 5 large cohorts of older subjects in studies from Europe and North America, we found that increasing intake of fish was modestly associated with slower rates of cognitive decline, in particular episodic memory decline—an early marker of AD risk. We did not find any suggestion of interactions between fish intake and AD-associated genotypes in this large sample.
Greater fish intake was related to a slower rate of global cognitive decline in our pooled analyses, although effect sizes appeared more modest than those we observed with episodic memory. Previous trials have generally provided null results on global cognitive function, perhaps due to shorter follow-up than our observational studies (21). Moreover, a number of observational studies reported relationships of fish intake/LCn-3PUFAs to change in global cognition or in nonmemory cognitive domains partly included in our global cognitive outcome (e.g., processing speed or verbal fluency) (6, 22–24). However, inconsistent findings have been reported (25, 26), and overall, results on global cognition have been mixed. It is possible that fish intake has a modest relationship with global cognition, a broad construct, and that our large sample size is needed to detect associations. Nonetheless, even modest associations can have a large public health impact. For example, one projection model estimated that delaying AD onset by only 1 year with preventive factors (e.g., fish) could reduce worldwide prevalence of AD by more than 9 million cases through 2050 (4).
Prospective studies on fish and episodic memory decline have been scarce. The association between increasing fish intake and lower rate of episodic memory decline that we found in this pooled analysis is consistent with individual results from the MAP cohort, although associations of fish to memory decline appeared stronger among APOEε4 carriers in MAP data (24). Our findings are also consistent with 2 small trials that showed a significant cognitive benefit of LCn-3 PUFA (using a dose equivalent to the one provided by 4 weekly servings of fish (approximately 1 g/day)). In healthy, young adults with habitually low DHA intakes, 6 months of DHA supplementation (approximately 1 g/day) improved memory but no other cognitive domain (12) (although in our large study, we also found significant tests of trend for the relationship of fish intake to global cognition). Likewise, in the Memory Improvement with DHA Study (MIDAS), 6 months of DHA supplementation (900 mg/day) in individuals aged ≥55 years with cognitive complaints improved memory but no other cognitive system (13).
In previous studies, 4 of the 5 cohorts (3C study, CHAP, MAP, and WHS) analyzed in the present meta-analysis reported slower decline in 1 or more cognitive domains with 1 or more seafood meals consumed per week. This is in contrast to findings from our pooled analyses, in which the apparent benefit was observed at a higher level of 4 or more meals consumed per week. At the same time, we did find statistically significant trends, such that each increasing unit of fish intake was associated with steadily decreasing rate of cognitive decline. This trend has not, to our knowledge, been reported in previous research, although in epidemiology, such trends are generally considered biologically plausible (27). One difference between the pooled analyses and these previous studies is that we used a subset of the cohort participants; this was due to our focus on interactions between fish intake and genetic factors as well as the desire to have comparable results in analyses with and without genetic variables. Our sample sizes within each study remained large after exclusions (e.g., in CHAP, we included 943 participants with both diet and genetic data available, of a possible 1,201 participants). Another difference is the inclusion of only reported white participants in the current pooled analysis, because there were modest numbers of minority participants across the 5 cohorts.
Our finding of stronger associations for episodic memory is plausible. Fish is the primary source of LCn-3PUFA, which may be particularly related to preservation of memory-related brain regions (i.e., the hippocampus). Animal studies have demonstrated that LCn-3 PUFA preferentially accumulates in the hippocampus and is involved in neuronal transmission, neurogenesis, synaptic plasticity, learning, and memory (14). In the 3C study, an association between higher plasma LCn-3PUFA and lower atrophy of the right medial-temporal lobe was found (28). All these findings provide strong mechanistic support for the observed relationship between LCn-3PUFA and episodic memory.
When we explored whether the relationship of fish to cognition could be modified by the genetic predisposition to AD, we did not find evidence of fish × gene interaction on change in either global cognition or episodic memory in a subset of well-replicated AD SNPs. Interactions between fish or LCn-3 PUFA and APOEε4 have been observed in a few previous studies; however, the studies were all of modest sample size, and no consistent associations have been reported (29). Indeed, inverse relationships between fish/LCn-3 PUFA and cognitive outcomes have been found limited to APOEε4 noncarriers (15, 30) or to APOEε4 carriers (24, 31), and most studies that investigated effect modification by APOEε4 carrier status did not identify any significant interaction (29).
An important strength of this pooled analysis is its ability, by combining 5 cohorts totaling more than 20,000 individuals, to identify modest associations and to examine genetic interactions; we included most of the large cohort studies with data on fish intake, repeated measures of cognition, and genetic data, and we adjusted for a large range of potential confounders. However, our study also had limitations. As discussed, a subset of the study participants was included in some of the cohorts, and the results cannot be generalized to nonwhite individuals. Also, as in any pooled analysis, there was inherent between-study variability in ascertainment of both fish intake and cognition. For example, the US cohorts used very similar FFQs and cognitive batteries, yet there was variability in the type of fish ascertained in the FFQs, and the tools used for ascertainment of both fish intake and cognitive function were different in the French 3C study (the FFQ was more limited, and episodic memory was based on single test subscore). However, the validity of the 3C study data on fish and episodic memory has been established previously (see Web Appendix 2). The cognitive tests used and methods of assessment may have differed in their sensitivity to measuring cognitive change; however, the findings appeared generally consistent across cohorts, and, specifically, most results from the 4 US cohorts did not materially differ from the 3C study or each other. Although our pooled sample included both sexes, the large majority of participants were female, precluding analyses of effect modification by sex; furthermore, with generally high levels of education, the sample was not ideally suited to investigate effect modification by education (although confounding by education is naturally limited in the cohorts, a potential strength). Finally, as in any observational study, confounding is an issue, although we adjusted for multiple potential confounders; residual confounding may persist, and findings should be interpreted with caution.
In summary, in this pooled analysis including 5 large cohorts of older persons with diet and cognitive data, we found significant trends of increasing fish intake associated with decreasing rates of cognitive decline, for episodic memory in particular. We found no evidence of effect modification by AD genes, indicating that any relationships of fish to cognition are likely similar regardless of genetic predisposition to AD.
Supplementary Material
AKCNOWLEDGMENTS
Author affiliations: Bordeaux Population Health Research Center, Institut National de la Santé et de la Recherche Médicale, U1219, and University of Bordeaux, Bordeaux, France (Cécilia Samieri, Jean-François Dartigues, Christophe Tzourio); Department of Internal Medicine, Rush University Medical Center, Chicago, Illinois (Martha-Clare Morris); Department of Behavioral Sciences, Department of Neurology and the Rush Alzheimer’s Disease Center, Rush University Medical Center, Chicago, Illinois (David A Bennett); Neuropsychiatrie: Recherche Epidémiologique et Clinique, Institut National de la Santé et de la Recherche Médicale, U1061, and Université de Montpellier, Montpellier, France (Claudine Berr); Institut Pasteur de Lille, Institut National de la Santé et de la Recherche Médicale, U1167, and Université Lille-Nord de France, Lille, France (Philippe Amouyel); Division of Preventive Medicine, Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts (Daniel I Chasman); Channing Division of Network Medicine, Department of Medicine, Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts (Francine Grodstein); and Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, Massachusetts (Francine Grodstein).
The Three-City study is conducted under a partnership agreement between Sanofi-Aventis, the Institut National de la Santé et de la Recherche Médicale (INSERM), and the Institut de Santé Publique et de Développement of the Victor Segalen Bordeaux 2 University. The Fondation pour la Recherche Médicale (FRM) funded the preparation and initiation of the study. The Three-City study is also supported by the Caisse Nationale Maladie des Travailleurs Salariés (CNAMTS), Direction Générale de la Santé (DGS), Mutuelle Générale de l’Education Nationale (MGEN), Institut de la Longévité, Regional Councils of Aquitaine and Bourgogne, Fondation de France, and Ministry of Research—INSERM Program “Cohortes et collections de données biologiques.” The 10-year follow-up was funded by grants from the ANR 2007LVIE 003 and the Fondation Plan Alzheimer. Genotyping of Three-City study participants was supported by the Fondation Plan Alzheimer, the Institut Pasteur de Lille, and the Centre National de Génotypage. The Nurses’ Health Study was funded by grants from the US National Cancer Institute (P01 CA87969), with additional support for genotyping from Merck Research Laboratories. The Women’s Health Study was supported by the US National Institutes of Health (grants HL043851, CA047988, HL080467, and AG015933), with collaborative scientific support and funding for genotyping provided by Amgen. The Chicago Health and Aging Project was supported by the National Institutes of Health (grants AG11101, AG13170, and AG10161). The Rush Memory and Aging Project was funded the National Institutes of Health (grants R01AG031553, R01AG17917 and RF1AG16819). C.S. was supported by a grant from the Fondation Plan Alzheimer.
Conflict of interest: none declared.
Abbreviations
- 3C study
Three-City study
- AD
Alzheimer disease
- APOE
apolipoprotein E
- CHAP
Chicago Health and Aging Project
- DHA
docosahexaenoic acid
- FFQ
food frequency questionnaire
- LCn-3 PUFA
long-chain omega-3 polyunsaturated fatty acid
- MAP
Rush Memory and Aging Project
- NHS
Nurses’ Health Study
- SNP
single nucleotide polymorphism
- WHS
Women’s Health Study
REFERENCES
- 1. Prasad MR, Lovell MA, Yatin M, et al. Regional membrane phospholipid alterations in Alzheimer’s disease. Neurochem Res. 1998;23(1):81–88. [DOI] [PubMed] [Google Scholar]
- 2. Söderberg M, Edlund C, Kristensson K, et al. Lipid compositions of different regions of the human brain during aging. J Neurochem. 1990;54(2):415–423. [DOI] [PubMed] [Google Scholar]
- 3. Lukiw WJ, Cui JG, Marcheselli VL, et al. A role for docosahexaenoic acid-derived neuroprotectin D1 in neural cell survival and Alzheimer disease. J Clin Invest. 2005;115(10):2774–2783. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Brookmeyer R, Johnson E, Ziegler-Graham K, et al. Forecasting the global burden of Alzheimer’s disease. Alzheimer’s Dement. 2007;3(3):186–191. [DOI] [PubMed] [Google Scholar]
- 5. Barberger-Gateau P, Letenneur L, Deschamps V, et al. Fish, meat, and risk of dementia: cohort study. BMJ. 2002;325(7370):932–933. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Morris MC, Evans DA, Tangney CC, et al. Fish consumption and cognitive decline with age in a large community study. Arch Neurol. 2005;62(12):1849–1853. [DOI] [PubMed] [Google Scholar]
- 7. Schaefer EJ, Bongard V, Beiser AS, et al. Plasma phosphatidylcholine docosahexaenoic acid content and risk of dementia and Alzheimer disease: the Framingham Heart Study. Arch Neurol. 2006;63(11):1545–1550. [DOI] [PubMed] [Google Scholar]
- 8. Samieri C, Féart C, Letenneur L, et al. Low plasma eicosapentaenoic acid and depressive symptomatology are independent predictors of dementia risk. Am J Clin Nutr. 2008;88(3): 714–721. [DOI] [PubMed] [Google Scholar]
- 9. Lopez LB, Kritz-Silverstein D, Barrett Connor E. High dietary and plasma levels of the omega-3 fatty acid docosahexaenoic acid are associated with decreased dementia risk: the Rancho Bernardo study. J Nutr Health Aging. 2011;15(1):25–31. [DOI] [PubMed] [Google Scholar]
- 10. Huang TL. Omega-3 fatty acids, cognitive decline, and Alzheimer’s disease: a critical review and evaluation of the literature. J Alzheimers Dis. 2010;21(3):673–690. [DOI] [PubMed] [Google Scholar]
- 11. Mazereeuw G, Lanctot KL, Chau SA, et al. Effects of omega-3 fatty acids on cognitive performance: a meta-analysis. Neurobiol Aging. 2012;33(7):1482.e17–1482.e29. [DOI] [PubMed] [Google Scholar]
- 12. Stonehouse W, Conlon CA, Podd J, et al. DHA supplementation improved both memory and reaction time in healthy young adults: a randomized controlled trial. Am J Clin Nutr. 2013;97(5):1134–1143. [DOI] [PubMed] [Google Scholar]
- 13. Yurko-Mauro K, McCarthy D, Rom D, et al. Beneficial effects of docosahexaenoic acid on cognition in age-related cognitive decline. Alzheimers Dement. 2010;6(6):456–464. [DOI] [PubMed] [Google Scholar]
- 14. Cunnane SC, Plourde M, Pifferi F, et al. Fish, docosahexaenoic acid and Alzheimer’s disease. Prog Lipid Res. 2009;48(5):239–256. [DOI] [PubMed] [Google Scholar]
- 15. Barberger-Gateau P, Raffaitin C, Letenneur L, et al. Dietary patterns and risk of dementia: the Three-City cohort study. Neurology. 2007;69(20):1921–1930. [DOI] [PubMed] [Google Scholar]
- 16. Seshadri S, Fitzpatrick AL, Ikram MA, et al. Genome-wide analysis of genetic loci associated with Alzheimer disease. JAMA. 2010;303(18):1832–1840. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Hollingworth P, Harold D, Sims R, et al. Common variants at ABCA7, MS4A6A/MS4A4E, EPHA1, CD33 and CD2AP are associated with Alzheimer’s disease. Nat Genet. 2011;43(5):429–435. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Lambert JC, Heath S, Even G, et al. Genome-wide association study identifies variants at CLU and CR1 associated with Alzheimer’s disease. Nat Genet. 2009;41(10):1094–1099. [DOI] [PubMed] [Google Scholar]
- 19. Radloff LS. The CES-D scale: a self-report depression scale for research in the general population. Appl Psychol Meas. 1977;1(3):385–401. [Google Scholar]
- 20. Laird NM, Ware JH. Random-effects models for longitudinal data. Biometrics. 1982;38(4):963–974. [PubMed] [Google Scholar]
- 21. Dangour AD, Andreeva VA, Sydenham E, et al. Omega 3 fatty acids and cognitive health in older people. Br J Nutr. 2012;107(suppl 2):S152–S158. [DOI] [PubMed] [Google Scholar]
- 22. Dullemeijer C, Durga J, Brouwer IA, et al. n 3 fatty acid proportions in plasma and cognitive performance in older adults. Am J Clin Nutr. 2007;86(5):1479–1485. [DOI] [PubMed] [Google Scholar]
- 23. Qin B, Plassman BL, Edwards LJ, et al. Fish intake is associated with slower cognitive decline in Chinese older adults. J Nutr. 2014;144(10):1579–1585. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. van de Rest O, Wang Y, Barnes LL, et al. APOE ε4 and the associations of seafood and long-chain omega-3 fatty acids with cognitive decline. Neurology. 2016;86(22):2063–2070. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Eskelinen MH, Ngandu T, Helkala EL, et al. Fat intake at midlife and cognitive impairment later in life: a population-based CAIDE study. Int J Geriatr Psychiatry. 2008;23(7):741–747. [DOI] [PubMed] [Google Scholar]
- 26. Beydoun MA, Kaufman JS, Satia JA, et al. Plasma n-3 fatty acids and the risk of cognitive decline in older adults: the Atherosclerosis Risk in Communities Study. Am J Clin Nutr. 2007;85(4):1103–1111. [DOI] [PubMed] [Google Scholar]
- 27. Hill AB. The environment and disease: association or causation? Proc R Soc Med. 1965;58:295–300. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Samieri C, Maillard P, Crivello F, et al. Plasma long-chain omega-3 fatty acids and atrophy of the medial temporal lobe. Neurology. 2012;79(7):642–650. [DOI] [PubMed] [Google Scholar]
- 29. Barberger-Gateau P, Samieri C, Féart C, et al. Dietary omega 3 polyunsaturated fatty acids and Alzheimer’s disease: interaction with apolipoprotein E genotype. Curr Alzheimer Res. 2011;8(5):479–491. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Whalley LJ, Deary IJ, Starr JM, et al. n-3 fatty acid erythrocyte membrane content, APOE varepsilon4, and cognitive variation: an observational follow-up study in late adulthood. Am J Clin Nutr. 2008;87(2):449–454. [DOI] [PubMed] [Google Scholar]
- 31. Samieri C, Féart C, Proust-Lima C, et al. Omega-3 fatty acids and cognitive decline: modulation by ApoEε4 allele and depression. Neurobiol Aging. 2011;32(12):2317.e13–2317.e22. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.