Relationship between diet and plasma long-chain n-3 PUFAs in older people: impact of apolipoprotein E genotype

Abstract

The main risk factors for Alzheimer's disease, age and the ϵ4 allele of the APOE gene (APOE4), might modify the metabolism of n-3 PUFAs and in turn, their impact on cognition. The aim of this study was to investigate the association between dietary fat and plasma concentrations of eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA) in elderly persons, taking the APOE4 genotype into account. The sample was composed of 1,135 participants from the Three-City study aged 65 years and over, of whom 19% were APOE4 carriers. Mean plasma proportions of EPA [1.01%, standard deviation (SD) 0.60] and DHA (2.41%, SD 0.81) did not differ according to APOE4. In multivariate models, plasma EPA increased with frequency of fish consumption (P < 0.0001), alcohol intake (P = 0.0006), and female gender (P = 0.02), and decreased with intensive consumption of n-6 oils (P = 0.02). The positive association between fish consumption and plasma DHA was highly significant whatever the APOE genotype (P < 0.0001) but stronger in APOE4 noncarriers than in carriers (P = 0.06 for interaction). Plasma DHA increased significantly with age (P = 0.009) in APOE4 noncarriers only. These findings suggest that dietary habits, gender, and APOE4 genotype should be considered when designing interventions to increase n-3 PUFA blood levels in older people.

Long-chain n-3 PUFAs [eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA)] could potentially exert a protective effect against brain aging due principally to their vascular, anti-inflammatory, and neuroprotective properties (1). Accordingly, several epidemiological studies have observed an inverse association between habitual consumption of long-chain n-3 PUFAs or fish, their main dietary source, and risk of cognitive decline or dementia (1). However, this protective association may be limited to individuals who do not carry the ϵ4 allele of the APOE gene (APOE4), the main genetic risk factor for Alzheimer's disease (AD) (2). Indeed, in the French Three-City (3C) study, the risk of all-cause incident dementia was decreased by 40% (P < 0.01) in elderly APOE4 noncarriers consuming fish at least once a week, while a nonsignificant reverse association [hazard ratio (HR) = 1.28, P = 0.55] was observed in APOE4 carriers (3). A very similar protective association against dementia was observed in the Cardiovascular Health Study with a risk reduction of 40% and 46% among APOE4 noncarriers who ate fatty fish two to four times per week or more than four times, respectively, compared with less frequent fish consumers, whereas in APOE4 carriers HRs were close to one and not significant (4). Regarding blood biomarkers, two prospective studies have found an inverse association between concentrations of EPA plus DHA in plasma (5) or total n-3 PUFAs in erythrocyte membranes (6) and subsequent cognitive decline only in APOE4 noncarriers. Controversial results exist as well, because another analysis of the 3C study found an inverse association between plasma EPA or DHA and decline in visual working memory only in APOE4 carriers (7). Moreover, the association between risk of dementia or AD and EPA or DHA blood concentration was not modified by APOE4 status in other studies (8–10).

As shown by a recent meta-analysis (11), most large randomized controlled trials (RCTs) failed to show any significant impact of DHA or EPA supplementation on cognitive decline in elderly participants. However, a per-protocol subgroup analysis of an RCT of DHA supplementation in patients with mild to moderate AD found that those receiving DHA in the APOE4-negative group had a significantly lower decline on the cognitive subscale of the Alzheimer's Disease Assessment Scale over 18 months than those receiving a placebo (12). There was no effect of supplementation with DHA on cognitive function in APOE4-positive participants in that study.

Taken together, these data suggest that APOE genotype might modulate the relationship between dietary intake and blood concentration of EPA and DHA and, in turn, the impact of these n-3 PUFAs on cognitive decline and risk for AD (2). Human beings are poor DHA synthesizers (13, 14), thus their blood concentration should closely reflect dietary intake. However, little is known about the determinants of blood concentrations of EPA and DHA and their variation in large samples of free-living populations with a spontaneous diet, especially in elderly individuals (15). Dietary intake of EPA and DHA explained only 12% of the variability of the Harris index (EPA plus DHA) in blood cell membranes in adults at high cardiovascular risk (16), suggesting that other factors are responsible for a substantial proportion of variation in their blood levels, among which genetic determinants such as the APOE genotype might have a major role. Indeed, an experimental intervention showed that a supplement of n-3 PUFAs (1.9 g EPA plus 1.1 g DHA per day) for 6 weeks induced higher plasma concentrations of both EPA and DHA in APOE4 noncarriers than in carriers (17). However, this study used doses of EPA and DHA much higher than usual dietary intake. In addition, older individuals have higher plasma EPA and DHA concentrations in response to supplementation with long-chain n-3 PUFAs than younger adults (18) or after an oral dose of carbon-13 DHA (19), suggesting significant differences in n-3 PUFA metabolism in older age.

Thus, the aim of the present study was to analyze the association between intake of dietary fat and blood concentrations of EPA and DHA in French elderly community dwellers, while taking APOE genotype into account.

METHODS

Population and sample

This cross-sectional study was conducted in the Bordeaux sample of the 3C study, a prospective cohort study of vascular risk factors for dementia which included 2,104 community dwellers at baseline in 1999–2000. Noninstitutionalized individuals aged 65 years and over who lived in this French city were eligible for recruitment into the 3C study. Details on the study and baseline characteristics were described previously (20). The protocol of the 3C study was approved by the Consultative Committee for the Protection of Persons participating in Biomedical Research of the Kremlin-Bicêtre University Hospital (Paris). This research adheres to the principles of the Declaration of Helsinki and follows the uniform requirements for manuscripts submitted to biomedical journals. All participants gave their written informed consent. Baseline data collection included socio-demographic and lifestyle characteristics including dietary habits, symptoms and complaints, main chronic conditions, neuropsychological testing, a physical examination, and blood sampling. Fatty acid analyses were performed at baseline in 1,518 participants from Bordeaux who accepted a blood draw. We excluded 108 subjects with missing data for certain dietary variables and 254 for other missing covariates, including APOE genotype. We further excluded 21 participants who were diagnosed as demented at baseline following the standardized procedure described elsewhere (20). Thus the main study sample included 1,135 subjects.

Plasma fatty acids

Fasting blood samples were collected at baseline in these participants in heparinized vacutainers and centrifuged at 1,000 g for 10 min, after which erythrocytes were separated from plasma. Total lipids were extracted from plasma with 5 ml of hexane/isopropanol (3:2, by vol). Plasma fatty acid composition was determined from 2 ml of the lipid extract after transformation into isopropyl esters (21). The isopropyl esters were separated by capillary gas chromatography, as previously detailed (8). The results for each fatty acid were expressed as percentage of total fatty acids. Plasma triglyceride concentration was also measured.

Dietary assessment

A brief food frequency questionnaire (FFQ) was administered at baseline to assess the dietary habits of the participants for broad categories of foods. Categories used in the present study were fish (including seafood) and meat (including poultry) as the main dietary sources of EPA, DHA, and arachidonic acid (a long-chain n-6 PUFA), respectively. Frequency of consumption was recorded in six classes: never, less than once a week, once a week, two to three times per week, four to six times per week, or daily. This questionnaire was similar to that developed and validated for the French national dietary surveys (22) recommended for use in French population studies (23).

We also considered intake of vegetable oils as dietary sources of α-linolenic acid (ALA) and linoleic acids, precursors of the n-3 and n-6 long-chain PUFAs respectively. Participants indicated the dietary fats used at least once a week for dressing, cooking, or spreading among the following list: corn oil, peanut oil, sunflower or grape seed oil (grouped in a single category as n-6 rich oils), olive oil, canola oil, walnut oil, mixed oil, margarine. The association between margarine or mixed oil and plasma fatty acids was not assessed because of the great variability of their composition. Consumption of added fat was considered in three categories for each fat type: never, moderate (i.e., used at least once a week either for dressing or cooking, not both), or intensive (i.e., used at least once a week for both dressing and cooking) as in previous papers (24, 25). When the number of intensive users was too small for statistical analyses, they were grouped with moderate users. Because alcohol may modify the metabolism of PUFAs (26), we also considered total alcohol intake. The number of glasses of wine and other alcoholic beverages consumed per week was recorded and converted into grams of alcohol per day.

Dietary habits of the whole 3C and Bordeaux samples have been described in detail elsewhere (27, 28).

APOE

APOE genotyping was carried out at the Lille Genopole, France. Because of the small number of ϵ4/ϵ4 homozygotes (n = 12), APOE4 allele carrier status was considered dichotomously: presence of at least one ϵ4 allele (APOE4 carrier, n = 216) versus no ϵ4 allele (APOE4 noncarrier, n = 919).

Other variables

Socio-demographic information recorded at baseline included age, gender, educational level (in four classes, see Table 1), and income [in four predetermined classes expressed in euros (€), Table 1]. Weight and height were measured and body mass index (BMI) was computed as the weight/height² ratio expressed in kg/m². According to previous studies in older persons, we used three classes for BMI: <21 kg/m² [threshold used for risk of undernutrition (29)], 21 to <25 kg/m² (normal weight), 25 to <30 kg/m² (overweight without obesity), ≥30 kg/m² (obesity). Physical activity was defined as regular when doing sport regularly or having at least 1 h of active leisure or household physical activity per day.

TABLE 1.

Characteristics of the participants according to the presence of at least one APOE4 allele (N = 1,135)

	APOE4
Socio-demographic Characteristics	Yes (n = 216)	No (n = 919)	P
Age, y, mean (SD)	73.5 (4.63)	74.3 (4.84)	0.03
Male gender, %	36.7	42.7	0.06
Education, %			0.36
No or primary	33.8	31.7
Secondary	25.0	27.5
High school	20.4	23.9
University	20.8	16.9
Income, €, % in each class			0.42
<750	5.5	7.4
750–1,500	29.2	33.1
1,500–2,250	24.1	24.6
>2,250	34.3	29.8
Refused to answer	6.9	5.1
BMI, kg/m2, % in each class			0.25
<21	10.2	7.1
21–25	32.0	30.5
25–30	44.4	44.9
>30	13.4	17.5
Physical activity, % yes	30.6	31.9	0.71
Triglycerides, mg/dl, mean (SD)	1.11 (0.60)	1.10 (0.52)	0.63

Statistical methods

In this study, we used EPA and DHA proportions in plasma at baseline as dependent variables. Mean EPA and DHA proportions in plasma were compared across classes of food intake also recorded at baseline (FFQ), APOE genotype and other covariates by Student's t-tests or ANOVA. Each variable significantly associated with EPA or DHA proportion at P < 0.15 in these univariate analyses was entered as an explanatory variable in multivariate linear regression models with log (EPA) or DHA as dependent variables, respectively. Log (EPA) was used in regression analyses because the distribution of plasma EPA was skewed. As APOE4 genotype and age were the main variables of interest, they were forced into the multivariate models. All multivariate models were also adjusted for plasma triglycerides. Normality and homoscedasticity of the residuals were checked. In order to examine potential interactions between dietary variables and APOE genotype, we tested all interactions between significant dietary variables (at P < 0.05 in bivariate regression models) and the APOE4 genotype on plasma concentrations of log (EPA) and DHA. As in previous papers using the same data (7, 8), the APOE4 genotype was considered as a potential effect modification factor when the interaction term was significant at P < 0.10. If so, stratified analyses were performed according to APOE4 carrier status.

RESULTS

The sample was composed of 469 men and 666 women aged 74.1 years on average (SD 4.8). The proportion of participants having at least one APOE4 allele was 19% including 12 homozygote and 204 heterozygote carriers. Carriers of the APOE4 were slightly younger but did not differ significantly from the noncarriers for any of the other socio-demographic characteristics, physical activity, or triglyceride plasma concentration (Table 1).

Description of EPA and DHA concentrations in plasma

Mean proportions of fatty acids in plasma were 1.01% (SD 0.60) for EPA and 2.41% (SD 0.81) for DHA. These proportions did not differ significantly according to APOE genotype (Table 2). There was no significant association between age and plasma EPA. Plasma DHA increased slightly but significantly with age, with a large inter-individual variability (DHA = 1.37777 + 0.013827 × age, Pearson correlation coefficient r = 0.08, P = 0.006). Regarding other socio-demographic characteristics, there was no difference in EPA and DHA proportions according to gender (Table 2). Both EPA and DHA were higher with higher income while only EPA increased with educational level. EPA was also higher in those with regular physical activity. DHA but not EPA varied inversely with BMI. Both EPA (Pearson correlation coefficient r = −0.15, P < 0.0001) and DHA (r = −0.18, P < 0.0001) were inversely associated with plasma triglycerides.

TABLE 2.

Plasma EPA and DHA proportions according to the genetic and socio-demographic characteristics of the participants (N = 1,135)

Characteristics	N	EPA %, mean (SD)	Pa	DHA %, mean (SD)	Pa
APOE4 carrier			0.19		0.73
Yes	216	0.96 (0.55)		2.39 (0.85)
No	919	1.02 (0.61)		2.41 (0.80)
Age, y			0.30		0.05
<75	664	1.02 (0.59)		2.35 (0.80)
75–80	329	1.04 (0.65)		2.50 (0.82)
80–85	121	0.94 (0.53)		2.43 (0.84)
≥85	21	0.86 (0.44)		2.45 (0.76)
Gender			0.11		0.11
Women	666	1.04 (0.64)		2.44 (0.81)
Men	469	0.98 (0.54)		2.36 (0.81)
Education			< 0.001		0.12
No or primary	364	0.95 (0.54)		2.35 (0.78)
Secondary	307	1.00 (0.57)		2.44 (0.83)
High school	264	1.00 (0.54)		2.36 (0.79)
University	200	1.18 (0.77)		2.50 (0.86)
Income, €			0.05		0.04
<750	80	0.83 (0.51)		2.16 (0.63)
750–1,500	367	1.02 (0.62)		2.39 (0.79)
1,500–2,250	278	1.00 (0.60)		2.40 (0.83)
>2,250	348	1.06 (0.61)		2.45 (0.82)
Refused to answer	62	1.00 (0.51)		2.51 (0.96)
BMI, kg/m2			0.23		0.04
<21	87	1.05 (0.64)		2.49 (0.88)
21–25	349	1.00 (0.56)		2.44 (0.84)
25–30	509	1.04 (0.64)		2.42 (0.80)
>30	190	0.95 (0.55)		2.25 (0.74)
Physical activity			0.04		0.23
Yes	359	1.07 (0.68)		2.45 (0.79)
No	776	0.99 (0.56)		2.38 (0.86)

^aP values of Student's t-tests for comparisons of means of EPA or DHA proportions for variables with two classes, ANOVA otherwise.

Univariate associations between dietary variables and plasma EPA and DHA

As expected, plasma EPA and DHA were considerably and significantly higher with higher frequency of fish consumption (Table 3). Plasma DHA, but not EPA, was inversely associated with frequency of meat consumption. Plasma EPA was significantly higher with higher alcohol consumption but there was no significant association of DHA with alcohol. Regarding types of added fat, there was a borderline significant (P = 0.06) inverse association between frequency of consumption of n-6 rich oils (sunflower or grape seed oil) and plasma EPA. There was no significant association between plasma EPA and any other source of added fat. There was no significant association of DHA with any added fat except with walnut oil, a rarely consumed oil (2.3% of the sample) for which moderate or intensive use was associated with lower plasma DHA.

TABLE 3.

Plasma EPA and DHA proportions according to diet and added fat consumption (N = 1,135)

Food/fats Consumption	N	EPA %, mean (SD)	Pa	DHA %, mean (SD)	Pa
Fish, servings per week			<0.0001		<0.0001
<1	105	0.87 (0.39)		2.08 (0.68)
1	375	0.90 (0.53)		2.23 (0.75)
2–3	574	1.06 (0.63)		2.47 (0.79)
>3	81	1.40 (0.75)		3.12 (0.92)
Meat, servings per week			0.18		0.03
≤3	375	1.05 (0.65)		2.48 (0.84)
>3	760	1.00 (0.57)		2.37 (0.79)
Alcohol, g/day			0.04		0.19
0 (never or past users)	232	0.92 (0.47)		2.33 (0.77)
0–12	478	1.02 (0.66)		2.45 (0.83)
12–24	227	1.01 (0.55)		2.44 (0.81)
24–36	76	1.05 (0.68)		2.40 (0.78)
>36	122	1.12 (0.59)		2.29 (0.81)
Corn oil			0.26		0.32
None	1091	1.01 (0.61)		2.40 (0.82)
Moderate/intensive use	44	1.09 (0.48)		2.52 (0.68)
Peanut oil			0.93		0.91
None	935	1.02 (0.62)		2.40 (0.82)
Moderate use	130	1.00 (0.48)		2.42 (0.75)
Intensive use	70	1.00 (0.48)		2.37 (0.75)
Sunflower oil or grape seed			0.06		0.16
None	528	1.05 (0.58)		2.44 (0.84)
Moderate use	281	1.02 (0.58)		2.42 (0.80)
Intensive use	326	0.95 (0.65)		2.33 (0.77)
Olive oil			0.10		0.18
None	363	0.97 (0.62)		2.34 (0.76)
Moderate use	458	1.00 (0.57)		2.42 (0.82)
Intensive use	314	1.07 (0.63)		2.45 (0.86)
Canola oil			0.23		0.90
None	1106	1.01 (0.60)		2.40 (0.81)
Moderate/intensive use	29	1.11 (0.57)		2.39 (0.75)
Walnut oil			0.89		0.03
None	1109	1.01 (0.59)		2.41 (0.81)
Moderate/intensive use	26	1.08 (0.94)		2.09 (0.87)

^aP values of Student's t-test for the comparison of two-class variables, ANOVA for variables with several classes.

Multivariate associations between dietary variables and plasma EPA and DHA: interaction with APOE genotype

There was no significant interaction between APOE4 genotype and any of the dietary variables on plasma EPA (all P > 0.10). Thus, APOE4, age, and all dietary variables significantly associated with plasma EPA at P < 0.15 in univariate analyses were entered into a single multivariate linear regression model on log (EPA) adjusted for socio-demographic variables and plasma triglycerides (Table 4). APOE4 genotype and age were not significantly associated with plasma EPA. As expected, plasma EPA increased strongly and significantly with increasing number of servings of fish per week, but also with the daily quantity of alcohol consumed. Intensive use of n-6 oils was inversely associated with plasma EPA. Among socio-demographic variables, women had a significantly higher plasma EPA than men. The proportion of variance explained by the model (R-square) was 10.8%.

TABLE 4.

Factors associated with log (plasma EPA): multivariate analysis (N = 1,135)

Variables	Beta Coefficienta	CI95%	Pa
APOE4	−0.04	−0.12 to 0.04	0.31
Age, y	−0.0004	−0.007 to 0.006	0.89
Fish, servings per week			<0.0001
<1 (ref)	0	—
1	−0.008	−0.12 to 0.11
2–3	0.12	0.01 to 0.23
>3	0.38	0.22 to 0.53
Alcohol, g/day	0.004	0.002 to 0.006	0.0006
Sunflower or grape seed oil			0.02
None (ref)	0	—
Moderate use	−0.02	−0.10 to 0.06
Intensive use	−0.11	−0.18 to −0.03
Olive oil			0.85
None (ref)	0	—
Moderate use	−0.02	−0.10 to 0.05
Intensive use	−0.01	−0.10 to 0.07
Gender			0.02
Men (ref)	0	—
Women	0.09	0.01 to 0.16

CI_95%, 95% confidence interval.

^aMultivariate linear regression adjusted for education, income, physical activity and plasma triglycerides.

Regarding plasma DHA, there was a borderline significant interaction between frequency of fish consumption in two classes (at least weekly consumption vs. less than weekly) and APOE4 (P = 0.06).Thus separate multivariate analyses were run stratified according to the presence of the APOE4 allele (Table 5). Plasma DHA significantly increased with age in APOE4 noncarriers only. Frequency of fish consumption was significantly and positively associated with plasma DHA in both APOE4 carriers and noncarriers, but the association was stronger and increased more with the number of servings in noncarriers. Indeed, DHA concentration increased less with fish consumption in APOE4 carriers than in APOE4 noncarriers: in APOE4 carriers, only those consuming at least three servings of fish per week had significantly higher plasma DHA than non-fish consumers, the reference category. For a given level of fish consumption, the magnitude of the β coefficients was also smaller in APOE4 carriers compared with APOE4 noncarriers. However, there was no significant difference between mean plasma DHA in APOE4 carriers and noncarriers across levels of fish consumption (all P > 0.05). Unexpectedly, plasma DHA was higher in APOE4 carriers who ate at least three servings of meat per week, while an inverse borderline significant association was observed in noncarriers. There was no difference in plasma DHA according to gender when taking diet into account in these multivariate models. Walnut oil was not included in the multivariate models because of the small number of consumers (n = 26). The proportion of variance explained by the model (R-square) was 18.5% in APOE4 carriers and 14.4% in APOE4 noncarriers whereas the proportion explained by fish consumption (partial R²) was 6.2% and 8.3% respectively, indicating that variation in plasma DHA was less sensitive to fish consumption in APOE4 carriers. No other significant interaction between dietary variables and APOE4 genotype was detected for DHA (all P > 0.10).

TABLE 5.

Factors associated with plasma DHA proportions according to APOE4 carrier status (N = 1,135)

	APOE4 carriers (n = 216)					APOE4 noncarriers (n = 919)
Variables	N	Mean Plasma DHA (SD)	Beta Coefficienta	CI95%	Pa	N	Mean Plasma DHA (SD)	Beta Coefficienta	CI95%	Pa
Age, y			0.01	−0.01 to 0.04	0.25			0.01	0.003 to 0.02	0.009
Fish, servings per week					0.005					<0.0001
<1	19	2.36 (0.78)	0	—		86	2.02 (0.65)	0	—
1	75	2.20 (0.74)	−0.16	−0.58 to 0.24		300	2.24 (0.75)	0.26	0.07 to 0.44
2–3	112	2.46 (0.84)	0.13	−0.28 to 0.54		462	2.48 (0.77)	0.42	0.24 to 0.59
>3	10	3.05 (1.38)	0.77	0.12 to 1.41		71	3.13 (0.85)	1.05	0.80 to 1.29
Meat, servings per week					0.04					0.05
≤3	69	2.29 (0.88)	0	—		306	2.52 (0.83)	0	—
>3	147	2.43 (0.83)	0.26	0.02 to 0.51		613	2.35 (0.78)	−0.11	−0.21 to 0.0001
Gender					0.08					0.53
Men (ref)	77	2.21 (0.76)	0	—		392	2.39 (0.81)	0	—
Women	139	2.49 (0.88)	0.22	−0.03 to 0.47		527	2.42 (0.80)	0.03	−0.07 to 0.14

CI _95%: 95% confidence interval.

^aMultivariate linear regressions adjusted for education, income, BMI, and plasma triglycerides.

DISCUSSION

In these French elderly community dwellers, fish consumption was positively associated with plasma EPA and DHA in both APOE4 carriers and noncarriers, but the frequency of fish consumption required to observe a significant rise in plasma DHA was higher in APOE4 carriers (at least three servings per week) than in noncarriers (at least one serving per week) compared with non-fish eaters. Other dietary components were also associated with plasma EPA and DHA, including alcohol for EPA and meat for DHA, but in the latter case an interaction with APOE4 genotype was also observed.

Our results suggest that the metabolism of dietary DHA is probably different according to APOE4 status in older adults but few studies have investigated this effect. Most studies have focused on the impact of APOE genotype on blood lipids (i.e., cholesterol and triglycerides) but not fatty acids (30–32). The modifying effect of APOE genotype on DHA metabolism has biological plausibility (2, 33). Indeed, APOE plays a major role in lipid transport in the blood, where it contributes to the delivery and clearance of serum triglycerides and other lipids (33). DHA metabolism is disturbed in APOE4 carriers (34). APOE4 is associated with a decreased activity of lipoprotein lipase, inducing inefficient delivery of free fatty acids such as DHA (33). APOE4 is also associated with increased oxidative stress (33), hence an increased level of lipid peroxidation which may affect DHA more severely than EPA because of its higher number of double-bonds. However, why there is an interaction with APOE4 for plasma DHA but not for EPA remains unclear. The coefficient of variation (SD/mean) for plasma EPA (0.59) was higher than for DHA (0.34) in our sample, which is most likely due to biological differences across the group rather than analytical issues. Conversion from EPA to DHA requires a second desaturation at the Δ6 position and limited β-oxidation (35). This step may represent a locus for metabolic regulation that facilitates the control of DHA synthesis. A supplementation with 6 g/day of n-3 PUFAs induced relatively higher changes in the composition of plasma phospholipids for EPA than for DHA (36). Similarly, increased EPA+DHA consumption (1.5 g/day for 8 weeks) significantly reduced cumulative concentration of labeled EPA in plasma phosphatidylcholine relative to baseline, but not of DHA or other fatty acid compartments, in older men (37). However, these intervention studies used amounts of EPA and DHA much higher than those encountered in usual diets.

Regarding the effect of age, our results do not fully support the conclusion of Hodson, Skeaff, and Fielding that “the majority of changes in the fatty acid composition of tissue and blood lipids that have been reported to occur with age can in part be explained by differences in dietary fatty acid intake rather than age-related perturbations in fatty acid metabolism” (15). Indeed, age was modestly (β = 0.01) but significantly associated with higher DHA in plasma independent of dietary intake, but only in APOE4 noncarriers. Short-term administration of an EPA+DHA supplement is accompanied by a higher increase in plasma DHA in healthy older adults than in younger adults (18). The elderly also metabolize an oral dose of carbon-13-labeled DHA differently than young adults (19). In neither of these studies were participants separated according to APOE4 genotype. Unlike other studies (38, 39), we did not observe higher plasma EPA in older subjects, but our sample was limited to individuals already aged 65 and over.

An intriguing finding of our study is the positive association between alcohol intake and plasma EPA but not DHA. A similar finding was observed in other studies among men (40, 41), while among women both plasma EPA and DHA were associated with alcohol intake, especially wine (41). Contrarily to in vitro studies, in vivo isotope tracer studies in primates showed that moderate alcohol consumption might increase the biosynthesis of EPA and DHA from their precursor ALA, suggesting an activation of the desaturation-elongation pathway (42). In hamsters, ethanol increased plasma DHA, and irrespective of whether the predominant type of PUFA in the diet was n-6 or n-3 PUFA (26). Furthermore, endogeneous biosynthesis from ALA is much more efficient for EPA than DHA, which may explain why we evidenced an association between alcohol intake and plasma EPA only. Alcohol consumption can also increase the level of lipid peroxidation in tissues (42). In our sample, wine is the most frequent source of alcohol intake. Thus, some components of wine other than ethanol (e.g., polyphenols) might increase long-chain n-3 PUFA concentrations by protecting them against lipid peroxidation (43). We showed that regular fish consumers were also often moderate wine drinkers, thus the respective effects of each nutrient are difficult to separate (44).

Unexpectedly, we found a positive association between frequency of meat consumption and plasma DHA in APOE4 carriers. Indeed, most meats are rich in arachidonic acid and low in n-3 PUFAs, so a negative association would be expected. This paradoxical finding needs to be confirmed by further investigations.

Although few variables were independent predictors of plasma DHA in our multivariate models, taking the APOE genotype into account provided a higher proportion of variance in plasma DHA explained by the model (18.5% in APOE4 carriers and 14.4% in noncarriers) than in previous studies. In the Spanish study (16), dietary intake of EPA and DHA explained only 12% of the variability of the omega-3 index (EPA plus DHA) in whole blood of subjects at high risk for cardiovascular disease, not taking the APOE genotype into account.

The strengths of this study include its large population-based sample of free-living elderly on their spontaneous diet, which included a wide range of fish consumption, and for whom APOE genotype was known. Our sample included individuals living close to the Atlantic Ocean with 90% having at least one fish meal per week which may represent a fairly high exposure to dietary EPA and DHA. The mean proportion of EPA in plasma (1.0%) in the elderly sample of the Canadian study (39) was very similar to ours but their mean proportion of DHA (1.7%) was much lower. However, that sample was very small (n = 25) and their fish consumption was not reported.

Our results do not compare directly with those of another French study (45) which was conducted among 382 adult high seafood consumers who had a very high intake of EPA and DHA combined (1,220 mg/day on average), which is not representative of the general population. In contrast to the present study, the authors found that at intakes >200 g per week of EPA+DHA, i.e., 28.6 g/day, their concentration in erythrocyte membranes no longer increased suggesting a ceiling effect in the relationship. However, such an extremely high consumption of EPA and DHA was not observed in our sample. Moreover, APOE genotype was not taken into account in their study.

Despite stratification and adjustment for many covariates, the proportion of variance of plasma EPA and DHA explained by the models, especially by fish consumption, remained modest. These findings suggest that other factors explain the variability of EPA and DHA concentrations, beyond those taken into account in our study. In particular, there could be an impact of intake of some nutrients that modulate the desaturation-elongation chain of ALA to EPA and DHA, e.g., Fe, Zn, vitamin B6, or Cu, as well as nutrients that modulate the susceptibility to their peroxidation, e.g., Se and vitamins E, C, or others. Other genetic polymorphisms might also contribute to explain a substantial proportion of variance of blood DHA (46).

Moreover, measurement error in dietary surveys tends to decrease the proportion of variance explained by controlled factors. The FFQ did not estimate portion size but only frequency of intake of broad food categories and thus could not be used to estimate actual intakes of EPA and DHA. However, there was good agreement between frequency of fish intake in the FFQ and mean estimated amounts of EPA and DHA intake in the 24 h recall in a comprehensive dietary survey conducted at 2 year follow-up (28), suggesting that the number of fish servings per week was an appropriate proxy for assessing dietary intake.

Plasma n-3 PUFA concentrations reflect only short-term consumption while proportions of fatty acids in cell membranes are thought to better represent usual intake, although the fatty acid composition of erythrocytes is also subject to rapid change (15). However, the good correlation observed between erythrocyte and plasma proportions of EPA and DHA in a subsample of 320 participants in the COGINUT-3C substudy (data not shown) suggests that plasma provides a good estimate of long-chain n-3 PUFA status.

CONCLUSION

These findings suggest that in an elderly population, individual characteristics such as gender, dietary habits, and also APOE genotype may modify the metabolism of EPA and DHA and possibly limit their bioavailability, which, in turn, may have practical implications when designing clinical trials with n-3 PUFA supplements. In the future, these results might also contribute to deriving specific dietary recommendations taking these characteristics into account with a perspective of “personalized nutrition”. However, the potential benefits for health outcomes of such personalized recommendations for fish and EPA/DHA intake still have to be evaluated. Further research is needed to better understand the determinants of n-3 PUFA concentrations in plasma of older individuals, including the role of genetic polymorphisms, in order to improve nutritional interventions in elderly people.