Associations between Serum Omega-3 Fatty Acid Levels and Cognitive Functions among Community-Dwelling Octogenarians in Okinawa,Japan: The KOCOA Study

Study design and participants

Data came from a cross-sectional study, part of a prospective pilot cohort study of community-dwelling adults aged 80 years and older living in Okinawa, Japan called the Key to Optimal Cognitive Aging Project (KOCOA). A detailed description of the recruitment process has been described elsewhere [26, 27]. Briefly, researchers visited 21 senior centers, explained the study protocol, and asked them to participate in the study. A request to join the study was made at the conclusion of each presentation. We recruited community-dwelling adults aged 80 years and older who needed no or partial assistance for all instrumental activities of daily living. The baseline recruitment occurred in 2007 and participants were assessed annually until 2011. Between November 2011 and April 2012, new participants were recruited using the same methods as before. These participants and those followed from the original cohort were used in the current analyses. Out of 191 participants (60 from the original cohort) who consented to the study and completed a face-to-face interview, four were taking anticholinesterase agents and two had frank dementia defined by Clinical Dementia Rating Scale (CDR) [28]≥1 and thus were excluded from the current study. The remaining 185 participants (CDR = 0 or 0.5) were assessed for their fatty acid levels using serum samples.

This study was approved by the Ethics Committee of the University of the Ryukyus, and the Institutional Review Board at Oregon Health & Science University. Informed consent was obtained from all participants prior to enrollment in the study.

Cognitive measures

Cognitive function was assessed by a trained interviewer using the Japanese version of the Mini-Mental State Examination (J-MMSE) [29, 30]. The J-MMSE is a measure of global cognitive function and ranges from 0 to 30 points, with higher scores representing better cognitive function. It takes approximately 10 minutes to administer and includes questions on orientation to time and place, attention and calculation, recall, language, and visual construction. The Verbal Fluency Letters (VFL) test was used to measure executive function. This test requires the participant to generate as many words as they can beginning with the letter “Ka” (in Japanese) in 60 seconds, and has been validated as an equivalent test to Initial Letter Fluency [31].

Serum fatty acid measurements

Venous blood samples were collected in the morning after an overnight fast. Serum was separated by centrifugation at 3000 r.p.m. for 15 min and stored at –80°C until analysis. The laboratory analyses were performed by SRL Inc. (Tokyo, Japan), using standard laboratory protocols.

PUFA estimates from food intake questionnaire

In our preliminary analysis, we found that serum DHA levels are negatively associated with age. To shed light on whether this is due to age-associated changes in dietary patterns, we examined the association between the amount of DHA and EPA obtained from food using a brief self-administered diet history questionnaire (BDHQ). The measurement is validated and explained in detail elsewhere [32, 33]. Brieflly, the BDHQ is a 4-page fixed-portion questionnaire that asks about the consumption frequency of selected foods to estimate the dietary intake of 58 food and beverage items during the preceding month. The BDHQ consists of five sections: (1) intake frequency of food and nonalcoholic beverage items; (2) daily intake of the Japanese staple rice and miso soup (typical Japanese soup); (3) frequency and amount of alcoholic beverage intake; (4) usual cooking methods, e.g., fried, steamed, or broiled in Japanese or Western Style; and (5) general dietary behavior, e.g., speed of eating and thickness of seasoning. The amount of DHA and EPA consumed from foods is estimated by using the information from these five categories and a validated algorithm.

Potential confounders and covariates

First we assessed the association between age and PUFA levels. Then we examined the association between PUFAs and cognitive functions, controlling for the following confounders: Age, hypertension, HbA1c, LDL Cholesterol, and obesity. Hypertension was defined by systolic blood pressure≥140 mmHg or diastolic blood pressure≥90 mmHg, or currently taking antihypertensive medication. The HbA1c value was estimated as a Japan Diabetes Society (JDS) method equivalent value (%), which was calculated by the formula: HbA1c (%) = HbA1c (%) (National Glycohemoglobin Standardization Program (NGSP) - 0.4% [34]. High HbA1c was defined by HbA1c level≥6.1% (JDS). High LDL-Cholesterol (LDL-C) was defined by LDL-C level≥140 mg/dL. Obesity was defined as Body Mass Index≥25 kg/m² according to guidelines published by the Japan Society for the Study of Obesity [35].

Statistical analysis

Associations between nutrient markers (EPA, DHA, AA, EPA/AA, DHA/AA, DHA + EPA) and age were examined by linear regression models. We then categorized J-MMSE and VFL into two groups and used t-tests to compare levels of nutrient markers by dichotomized J-MMSE (J-MMSE <24 versus J-MMSE≥24) and dichotomized VFL (VFL <5 versus VFL≥5). There is no predefined clinical threshold for VFL. Therefore, we dichotomized below/above the lowest 33% as a cutpoint (score of 5). To address the skewed distributions, we used a log-transformation of nutrient markers for the above univariate analyses except DHA which was normally distributed. Our preliminary results showed no association between nutrient markers and VFL. Therefore, subsequent multivariate analyses were conducted only for J-MMSE. Nutrient markers which were significant on univariate analyses were included in the multivariate models. We divided J-MMSE into three groups (Group 1: Lowest tertile, Group 2: Middle tertile, Group 3: Upper tertile). Those in the lowest tertile of J-MMSE score were compared to participants with higher tertiles using cumulative logit models (i.e., the model compares (a) Group 1 versus 2 and above, and (b) Groups 1 and 2 versus Group 3 in outcomes). This model generates two intercepts and one coefficient, assuming the effects of independent variables on dependent variables are the same for the comparisons of (a) and (b) above. Our preliminary analysis confirmed that this model assumption was met in our data. We divided each nutrient marker into three groups: Lowest 25%, middle 50%, and highest 25%, with the lowest 25% group being the reference group and included each marker separately into models controlling for age, years of education, gender, obesity, hypertension, high HbA1c, and high LDL-C, with the latter five items as categorical variables. All statistical analyses were performed with JMP^® 11 for Windows (SAS Institute Inc., Cary, NC, USA). Statistical significance was set at p <0.05. Additionally, we provided the false discovery rate approximation adjusted p-value in table footnotes as multiple comparion adjusted type I error rates, using equation α(m + 1)/2m, where α is 0.05 and m is the number of tests.

Associations between PUFAs and age

Kuriki et al. found that plasma EPA and DHA increased with age among Japanese female dietitians under 66 years old [37]. Otsuka et al. also reported that omega-3 PUFAs including α-linoleic acid, EPA, DHA, and docosa-pentaenoic acid increased with age, whereas omega-6 fatty acids including linoleic acid and AA decreased with age among those 40 to 88 years [38]. Authors in both studies suggested changes in diet (older adults may eat more fatty fish and less red meat [39]) as a possible reason for this age trend. The participants in the above two studies were younger with a wider age range than those in our study. Using autopsy data from participants between 33 and 92 years old, the level of in vivo omega-3 PUFAs, particularly DHA, was found to decrease with age [40]. In our study, the consumption of EPA and DHA estimated from a food intake questionnaire was stable regardless of age. EPA is transformed to DHA by a converting enzyme. Reduced ability to convert the enzyme or to absorb DHA might be reasons for the age-associated decline in DHA levels seen in our study. Alternatively, age-related defects in antioxidant systems may increase lipid peroxidation and thereby decrease levels of omega-3 PUFAs [41]. Further studies are warranted to confirm the underlying mechanisms of the age-associated decline in DHA among the oldest old.

The levels of DHA, EPA, and EPA/AA ratio found in our cohort were similar to those found in several other Japanese studies; however, our cohort’s AA levels were higher in comparison [42, 43]. For example, in Yanagisawa et al’s study, the mean level of serum DHA was 123.2±27.0 μg/ml, EPA was 81.9±31.1 μg/ml, AA was 119.8±22.7 μg/ml, and EPA/AA ratio was 0.68±0.22 among those aged 65 and older. In Yagi et al’s study, the mean level of serum DHA was 132.5±52.2 μg/ml and EPA was 68.9±40.4 μg/ml. Our mean values were DHA: 129.2±36.3 μg/ml, EPA: 62.5±52.2 μg/ml, AA: 195.1±47.7 μg/ml, and EPA/AA ratio: 0.34±0.34 (Table 1). The reason our cohort has higher AA values may be due to the fact that Okinawa had been under the United States occupancy from the end of World War II until 1972 and therefore western food culture has been more prevalent, leading to higher consumption of red meat. As for comparisons with non-Japanese cohorts, using plasma samples, Bowman et al. [44] reported the mean DHA and EPA values were 68.1±17.8 μg/ml and 16.5±10.5 μg/ml, respectively, among community-dwelling elderly aged 80 and older (mean age of 85.7±10) with normal cognition in Portland (Oregon) in the United States. These values are much lower than those shown in our study. The mean EPA value reported in the study by Bowman et al. is lower than the 25% lowest cutpoint of EPA (33.5 μg/ml) in our study, with which we found significant association with lower cognitive test scores. Sekikawa et al. [45] examined serum PUFAs values among quadragenarians in Japan and the United States. They reported that mean serum percentage of PUFAs was much higher among the Japanese than in the United States (9.08% versus 3.84%, p <0.01), coinciding with our finding among the octogenarians. The implication of large differences in PUFA values across regions or countries on cognitive declines or dementia incidence would be an interesting future research area. As previously suggested, both inter- and intra- cultural examinations are required to give a full perspective on the association between nutrition and health [26].

The differential effect of DHA and EPA on cognition

We found that higher levels of serum EPA, but not DHA, were associated with better cognitive function as measured by J-MMSE. Studies have shown that omega-3 PUFAs, mainly DHA and EPA, have protective effects with reagrds to anti-inflammatory properties [46 –48], cardiovascular disease [45, 49], and cognitive decline. The Framingham Heart Study showed that higher plasma DHA levels were associated with a 47% reduction in the risk of developing all-cause dementia and a 39% reduction in AD over nine years of follow-up. Their mean cohort age at baseline was about 75 years old. The Framingham Study did not examine EPA and its effects on dementia incidence. The Three-City Study from Bordeaux (France) reported that higher plasma EPA levels, but not DHA levels, were associated with a lower incidence of dementia of all types (HR = 0.69; 95% CI: 0.48–0.98) [50]. Martins et al. examined the differential effect of EPA and DHA on depression and cognition through an interventional meta-analysis of 28 randomized controlled trials [51]. They reported that EPA had a larger effect than DHA on depressive status and cognition function. Umegaki et al. found that higher serum triglycerides (TG) were associated with cognitive decline in Japanese non-demented older adults [52], and omega-3 PUFAs, mainly EPA and DHA, were found to reduce serum TG concentrations [53, 54]. Harris et al. reported potential TG-lowering mechanisms of EPA and DHA via inhibition of hepatic very-low-density-lipoprotein cholesterol. In a review of omega-3 PUFA supplementation clinical trials to reduce cardiovascular risk, Itakura et al. reported that an EPA/AA ratio >  0.75 was associated with lower major coronary artery events in a Japanese cohort with mean age of 61 years (SD = 9) [55]. EPA may also have neuroprotective effects [56]; prevention of neuronal apoptosis prevents hippocampal dysfunction [57]. EPA was found to be more effective at reducing inflammation in vivo than DHA [58], although both dietary EPA and DHA incorporate into cell membranes. The excessive production of pro-inflammatory eicosanoids derived from AA are competitively compensated for by the anti-inflammatory products of EPA metabolism [59]. Interestingly, we found that the odds of having higher MMSE scores were similar regardless of being in the highest 25% or the middle 50% of EPA or EPA + DHA levels. No dose response relation beyond the lowest 25% threshold suggests that as long as octogenarians keep their levels higher than the 25% tile in EPA and EPA + DHA, they could lower their risk for worsening cognitive outcomes. The cut off levels for the lowest 25% tile for EPA and EPA + DHA were 33.5 μg/ml and 140.65 μg/ml, respectively.

Vitamin D: A potential confounder

Vitamin D has been found to be associated with cognitive functions. Most recently, Miller et al. reported that those with dementia had a lower level of serum 25-hydroxyvitamin D compared to those with MCI and normal cognition [60]. They also reported that the rates of decline in episodic memory and executive function over time were steeper among vitamin D-deficient participants than those with adequate status [61]. It is possible the effect of EPA on cognition found in our study might be, at least partly, due to vitamin D. Therefore, as a post hoc analysis, we included vitamin D consumption levels estimated from the BDHQ in the multinomial logit models. After controlling for vitamin D level, serum EPA maintained its significant association with J-MMSE scores. Although vitamin D itself was not associated with the J-MMSE scores, its levels estimated from the food intake questionnnarie in our study had a large variability, making inferences difficult. The examination of their protective effects on cognition among the oldest old is further warranted.

The association of DHA and EPA with executive function

Bowman et al. [10, 43] found significant associations between executive function and DHA as well as EPA. We did not find any associations between EPA, DHA, and executive function. Although the test we used to measure executive function in this study (Verbal Fluency starting with “Ka” (Table 3)) taps the executive domain, it also measures phonetic abilities (language-based executive functions), while Bowman et al. used Trail Making Test B. The difference in tests could explain the discrepancy in findings between our study and theirs.

Limitations of our study include its cross-sectional design and thus our inabilityto demonstrate a casual relationship between serum omega-3 PUFAs and cognitive function. Also we could not examine the combined effects of PUFAs with other nutrient markers such as trans fat, saturated fats, or vitamins B, C, D, and E, which could have synergistic effects on cognition. Moreover participants were healthy volunteers living in Okinawa, which may limit the generalizability of the study results.

In conclusion, we found that higher serum EPA and DHA + EPA levels are associated with better cognitive function among community-dwelling non-demented volunteers aged 80 years and older. Higher intake of fatty fish may promote better cognitive function, even in the relatively healthy oldest old population. Further research is needed to understand the association between circulating EPA and DHA and dementia.