ω-3 Ethyl ester results in better cognitive function at 12 and 30 months than control in cognitively healthy subjects with coronary artery disease: a secondary analysis of a randomized clinical trial

ABSTRACT

Background

Omega-3 (n–3) fatty acids have shown benefit in cognitively impaired subjects, but the effect on cognitively healthy older subjects is unclear.

Objectives

Our aim was to determine if long-term, high-dose ω-3 ethyl esters, EPA (20:5n–3) and DHA (22:6n–3), prevent deterioration of cognitive function in cognitively healthy older adults.

Methods

A total of 285 subjects with stable coronary artery disease (CAD) on statin treatment were randomly assigned to 3.36 g EPA and DHA or none (control) for 30 mo. Cognitive function was assessed in all 285 subjects at baseline and in 268 and 250 subjects who returned at 12- and 30-mo follow-up, respectively, with neuropsychological testing as a prespecified secondary outcome. A completer's analysis, along with a sensitivity analysis carrying forward the last observation, was performed.

Results

Over the 30-mo period, subjects randomly assigned to EPA and DHA had significantly better scores than control for verbal fluency, language, and memory (mean: 1.08; 95% CI: 0.25, 1.91; P = 0.011) and 2 tests of visual-motor coordination (mean: −2.95; 95% CI: −5.33, −0.57; P = 0.015 and mean: −9.44; 95% CI: −18.60, −0.30; P = 0.043, respectively). The better scores for EPA and DHA were due to an improvement at 12 mo compared with baseline in verbal fluency, language, and memory (P = 0.047) and 2 tests of visual-motor coordination (P = 0.033 and P < 0.001, respectively), whereas control had no change. Post hoc analyses indicated no difference by age, sex, or diabetes status.

Conclusions

Cognitively healthy older adults with stable CAD randomly assigned to high-dose EPA and DHA had improved cognitive function over a 30-mo period compared with control. These findings may be especially important for CAD patients because CAD is a risk factor for cognitive decline.

This trial was registered at clinicaltrials.gov as NCT01624727.

Introduction

Dementia is a major and increasing global health challenge, with an estimated 50 million people worldwide currently living with dementia according to the World Alzheimer Report 2018 (1). This number is expected to increase to 152 million by 2050. Worldwide annual dementia costs are $1 trillion. Cognitive impairment is particularly prevalent among the elderly, and its burden tends to increase with age. Approximately 15%–20% of those age 65 y and older have mild cognitive impairment which is an intermediate stage in the continuum from normal aging to dementia (2). Cardiovascular disease, specifically coronary artery disease (CAD), has been implicated as a risk factor for the development of dementia. CAD causes ischemia, thus reducing cerebral blood flow, which contributes to the development of dementia (3). Patients with a history of CAD have greater decline in global cognition, verbal memory, and executive function over an mean follow-up of 6.9 y than those without CAD (4). Therefore, methods to prevent decline are important in subjects with CAD.

Omega-3 (ω-3) fatty acids are PUFAs required in the diet to provide the long-chain EPA (20:5n–3) and DHA (22:6n–3) for metabolic structure and function. Brain tissue membranes are rich in DHA (5). Although EPA is almost undetectable in the brain, it is implicated in the reduction of depressive symptoms (6, 7). Both EPA and DHA reportedly have neuroprotective effects via mechanisms such as suppression of inflammation, regulation of neurogenesis, and protection against oxidative stress (7, 8). DHA comprises ∼30% of neuronal membranes (9). High dietary intake and plasma concentrations of DHA have been associated with reduced risk of dementia in the Rancho Bernardo study (10). Diets which are rich in foods high in ω-3 fatty acids have been associated with a lower risk of Alzheimer disease (11). Low serum concentrations of DHA have been reported in patients with Alzheimer disease compared with control values (12–14). Low erythrocyte concentrations of DHA have been associated with cognitive decline (15). Moreover, in Framingham Heart Study participants without clinical dementia, lower erythrocyte concentrations of DHA were associated with smaller brain volume and cognitive impairment (16). Supplementation with 900 mg DHA daily prevented cognitive decline in subjects with age-related cognitive decline (17). Furthermore, a meta-analysis reported that supplementation with DHA improved episodic memory in adults with mild memory complaints (5). Yet, despite several studies that have highlighted the benefit of ω-3 fatty acids in subjects with mild cognitive impairment or dementia, studies involving cognitively healthy older subjects have been limited in size and duration; therefore, the effect in subjects with normal cognitive function is unclear.

We conducted a randomized, parallel, controlled study of high-dose ω-3 fatty acids—EPA and DHA—over 30 mo. The primary outcome is effect of EPA and DHA on progression of coronary arterial plaque at 30 mo of follow-up and has been reported (18). As a prespecified secondary outcome, we report the effects on cognitive performance in patients with stable CAD measured using neuropsychological cognitive testing in the current report. Our hypothesis was that supplementation with high-dose EPA and DHA would prevent decline in cognitive function measured by standardized cognitive tests in patients with known CAD compared with subjects randomly assigned to no EPA and DHA.

Methods

Study design

The HEARTS (Slowing HEART diSease with lifestyle and omega-3 fatty acids) trial (NCT01624727) is a randomized parallel study of subjects with CAD conducted at Beth Israel Deaconess Medical Center, Boston, MA. The protocol was approved by the Beth Israel Deaconess Medical Center Institutional Review Board; all subjects signed informed consent.

Participants

Participants were aged 37–80 y at time of enrollment and had stable CAD defined as ≥1 of the following: ≥50% stenosis in ≥1 coronary artery at catheterization, previous myocardial infarction (≥6 mo prior), percutaneous coronary intervention (≥6 mo prior), coronary bypass surgery (≥12 mo prior), abnormal exercise treadmill test, or an area of reversible ischemia on nuclear imaging, pharmacologic stress testing, or stress echocardiography with subsequent revascularization. Exclusion criteria included history of memory loss, dementia, or Alzheimer disease.

Randomization and study drug

Randomization was computer-allocated in blocks of 4 and stratified by presence or absence of diabetes. Participants were randomly assigned to receive either open-label ω-3 ethyl esters or none (termed control). Subjects in the ω-3 ethyl ester group received 3.36 g ω-3 ethyl ester daily as 4 soft gels, each containing predominantly 465 mg EPA and 375 mg DHA, for a total daily dose of 1.86 g EPA and 1.5 g DHA for 30 mo (Lovaza; GlaxoSmithKline). Subjects returned bottles of ω-3 ethyl ester at each study visit and pill counts were done to assess compliance.

Outcomes and data collection

Prespecified secondary outcomes, which have been published, include prevention of progression of the urine albumin:creatinine ratio in diabetic subjects, prevention of pain, stiffness and joint replacement, and effect of systolic compared with diastolic blood pressure on coronary plaque volume (19–21). The current report is a prespecified secondary outcome to examine the effect of EPA and DHA on cognitive function over a 30-mo period. No power calculation was performed for this secondary analysis.

Cognitive tests

Cognitive function was assessed at baseline, 1 y, and 30 mo using the following 5 neuropsychological cognitive tests with details summarized in Table 1. The Mini-Mental Status Examination (MMSE) measures global cognition (score range: 0–30) and was used to document a cognitively healthy status at the baseline exam (22). The Controlled Oral Word Association Test (COWAT) 1 and 2 measures verbal fluency, language, and memory assessed by the number of words named over 2 min for 2 separate lists (23). For COWAT 1, we asked subjects to name animals and for COWAT 2, supermarket items. The Trail Making Test (TMT) parts A and B measures visual-motor coordination (time to complete in seconds) (24). The Digit-Symbol Substitution Test (DSST) measures psychomotor speed (number of symbols correctly matched with the corresponding digit in 120 s) from the Wechsler Adult Intelligence Scale-Revised (25). The Rey Auditory Verbal Learning Test (RAVLT) measures memory and recall with RAVLT1 measuring immediate recall, RAVLT2 measuring short delay, RAVLT3 measuring delayed recall, and RAVLT4 measuring delayed recognition (26, 27).

TABLE 1.

Descriptions of cognitive tests administered¹

Cognitive function test	Test assessment	Scoring
COWAT	Verbal fluency, language, and memory	Number of items from each category named in 120 s. COWAT 1 = animals; COWAT 2 = supermarket items. A higher score is better.
MMSE	Global cognition	Number of questions posing problems answered correctly out of a possible total of 30. A higher score is better.
TMT—Parts A and B (TMT A, TMT B)	Visual-motor coordination, attention, and ability to switch between tasks	Time taken for subject to first connect consecutively numbered circles (Part A) and then connect the same number of consecutively numbered and lettered circles alternating between the 2 sequences (Part B). A lower score is better.
DSST	Psychomotor speed, sustained attention, response speed, and visual-motor coordination	Number of symbols correctly matched with their corresponding digit in 120 s. A higher score is better.
RAVLT	Memory and recall: storage and retrieval of newly acquired verbal material	Average number of words recalled (0–15).RAVLT1: immediate recall as a sum of 4 trials. RAVLT2: short delay recall. RAVLT3: delayed recall. RAVLT4: delayed recognition.A higher score is better.

¹COWAT, Controlled Oral Word Association Test; DSST, Digit-Symbol Substitution Test; MMSE, Mini-Mental Status Examination; RAVLT, Rey Auditory Verbal Learning Test; TMT, Trail Making Test.

Statistical analyses

The results were examined by treatment assignment. Categorical variables were expressed as counts and percentages and compared using the chi-square test. Normality tests were conducted using the Shapiro–Wilk test. Continuous variables were reported as mean ± SD for normally distributed variables and as median [IQR] for variables which were not normally distributed. Continuous variables were compared using either paired or unpaired t tests for normally distributed variables, the Wilcoxon–Mann–Whitney U test for nonnormally distributed variables, and Wilcoxon's signed-rank test for paired analyses of nonnormally distributed variables. A linear mixed-effects repeated-measures model was fitted to the data for each cognitive test. Test score was the dependent variable; randomized treatment, age, sex, and diabetic status were fixed effects; baseline score was a covariate; assessment time (12 or 30 mo) was a repeated measure; and subject was a random effect. This model has the advantage of adjusting scores at 12 and 30 mo for any imbalances between the groups in baseline score and for any differences in age, sex, and diabetes status which may influence scores at 12 and 30 mo. Interactions between baseline variables and randomized treatment were added to the model, whereas consistency of the treatment effect between 12 and 30 mo was assessed by adding treatment-by-assessment time as an interaction term. Post hoc exploratory analyses were conducted to examine the effect of EPA and DHA stratified by median age, sex, and diabetes status. Logistic regression models were used to determine whether missed assessments could have introduced bias. The models had missing/nonmissing assessment as a binary outcome and included factors for diabetes, sex, age, randomized treatment, and scores for 9 cognitive tests at the previous assessment. A sensitivity analysis using last observation carried forward for subjects who withdrew from the study between 12 and 30 mo was also performed. A 2-sided P value <0.05 was considered statistically significant. Statistical analyses were performed using SPSS version 20.0 (IBM Corp.) and the R programming language.

Results

A total of 285 subjects with stable CAD were randomly assigned to 3.36 g EPA+DHA or no EPA+DHA (control) for 30 mo. A total of 268 subjects returned at 1 y and 250 returned at 30 mo for cognitive testing. Supplemental Figure 1 shows the Consolidated Standards of Reporting Trials (CONSORT) diagram and the reasons for the dropouts. Our logistic regression analyses indicated that characteristics were not different in those missing the 12-mo or the 30-mo assessment. In view of the large number of factors examined and the low number of patients withdrawing prematurely, our analyses suggest that subjects with missing data may not be substantially different from those without missing data, and it is valid to assume that missing data are missing at random and have not introduced bias.

As Table 2 shows, no significant differences were observed in baseline characteristics between those who returned at 12 mo in the EPA+DHA and control groups. Mean ± SD age was 63.4 ± 7.5 y for the total group (83% male). All subjects had a score of ≥29 on the MMSE at baseline and were deemed cognitively healthy.

TABLE 2.

Baseline characteristics according to treatment assignment¹

	EPA+DHA (n = 138)	Control (n = 130)	P value
Demographic characteristics
Age, y	62.9 ± 7.6	63.9 ± 7.4	0.27
Male sex	115 (83.3)	108 (83.1)	0.82
Diabetics	39 (28.3)	40 (30.8)	0.81
Anthropometrics and blood pressure
Weight, kg	91.8 ± 13.6	90.2 ± 15.2	0.48
BMI,2 kg/m2	30.8 ± 3.6	30.6 ± 3.7	0.66
Waist, cm	106.9 ± 10.3	106.7 ± 11.2	0.71
Systolic BP, mm Hg	125.1 ± 14.6	124.4 ± 14.8	0.88
Diastolic BP, mm Hg	73.1 ± 10.0	72.4 ± 8.9	0.84
Biochemical profile
HbA1c, %	6.1 ± 0.8	6.3 ± 1.1	0.27
Glucose, mg/dL	104.8 ± 27.4	108.6 ± 37.3	0.29
Lipid concentrations,3 mg/dL
Total cholesterol	152.5 ± 37.5	152.1 ± 34.0	0.76
LDL-C	77.7 ± 26.9	78.2 ± 28.3	0.35
HDL-C	47.2 ± 14.8	46.8 ± 14.4	0.72
Triglyceride	121.5 [80.3–178.8]	114.0 [78.0–161.0]	0.44
Inflammatory markers
hs-CRP, mg/L	1.0 [0.4–2.5]	1.0 [0.4–3.1]	0.08
WBC, 109 cells/L	6.6 ± 2.0	6.6 ± 1.6	0.59
Monocytes, cells/μL	542.3 ± 173.0	516.9 ± 170.7	0.16
Neutrophils, cells/μL	4238.9 ± 1767.5	4192.4 ± 1341.1	0.74
Lymphocytes, cells/μL	1629.9 ± 557.6	1694.6 ± 632.6	0.23
Medications
Statin therapy	133 (96.4)	123 (94.6)	0.37
Aspirin	133 (96.4)	124 (95.4)	0.53

¹Values are mean ± SD, n (%), or median [IQR]. Categorical variables (male sex and diabetes) were expressed as n (%) and compared using the chi-square test. All normally distributed continuous variables were compared using unpaired t tests except for triglyceride and hs-CRP, which were compared with the Wilcoxon–Mann–Whitney U test because they were nonnormally distributed variables. BP, blood pressure; HbA1C, glycated hemoglobin; HDL-C, HDL cholesterol; hs-CRP, high-sensitivity C-reactive protein; LDL-C, LDL cholesterol; WBC, white blood cell.

²Calculated as weight in kilograms divided by height in meters squared.

³Multiply by 0.02586 to convert cholesterol values to mmol/L and by 0.01129 to convert triglyceride to mmol/L.

Effect of treatment with EPA and DHA compared with control

Baseline cognitive function scores were well matched between the EPA+DHA and control groups (Table 3). Table 3 shows point estimates for the mean (95% CI) values from the repeated-measures model of effects of treatment with EPA and DHA compared with control. In those randomly assigned to EPA and DHA, scores on verbal fluency, language, and memory (COWAT 1, P = 0.011) and 2 tests of visual-motor coordination—the Trail Making Test parts A and B (P = 0.015 and 0.043, respectively)—were significantly better than control over the 30-mo period. P values for interaction between treatment and assessment time were not statistically significant which suggests that the differences in cognitive scores between the EPA+DHA and control groups were similar at 12 and 30 mo. We carried out sensitivity analyses using “last observation carried forward” for subjects who withdrew from the study between 12 and 30 mo. Results from the repeated-measures mixed-effects model were numerically very similar where the treatment effect for last observation carried forward for COWAT 1 was 1.06 (P = 0.011) and for TMT A was −2.94 (P = 0.015). The only exception was for TMT B where the treatment effect of −9.44 (P = 0.043) for the main analysis in Table 3 was −9.01 (P = 0.052) in the sensitivity analysis. Bearing in mind that last observation carried forward analyses are generally regarded as conservative, we consider the results of our primary analysis shown in Table 3 as robust with regard to missing data.

TABLE 3.

Summary of scores for cognitive tests and mean (95% CI) values for effects of treatment with EPA and DHA compared with control over the 30-mo period¹

	Baseline		12 mo		30 mo		Effect of treatment (EPA+DHA − control)		P value for interaction (treatment × time)
Cognitive function test	EPA+DHA (n = 144)	Control (n = 141)	EPA+DHA (n = 138)	Control (n = 130)	EPA+DHA (n = 129)	Control (n = 121)	Mean (95% CI)	P
COWAT 1	20.4 ± 5.4	19.6 ± 4.7	21.2 ± 5.0	19.5 ± 5.0	20.9 ± 5.2	19.2 ± 5.5	1.08 (0.25, 1.91)	0.011	0.22
COWAT 2	26.2 ± 5.8	26.2 ± 5.8	26.1 ± 6.2	25.7 ± 5.8	25.7 ± 5.7	24.9 ± 6.2	0.50 (−0.43, 1.42)	0.29	0.18
DSST	58.6 ± 13.5	57.2 ± 13.8	61.5 ± 13.5	57.7 ± 14.9	61.4 ± 13.5	58.7 ± 15.0	1.32 (−0.32, 2.96)	0.12	0.79
TMT A2	35.2 ± 13.7	35.8 ± 13.5	33.0 ± 11.9	36.8 ± 11.0	34.0 ± 12.6	38.1 ± 18.0	−2.95 (−5.33, −0.57)	0.015	0.08
TMT B2	77.5 [60.4–101.2]	80.2 [60.7–116.2]	74.0 [60.4–95.9]	80.1 [64.5–111.3]	74.5 [60.0–101.9]	80.2 [62.5–119.1]	−9.44 (−18.60, −0.30)	0.043	0.08
RAVLT1	31.2 ± 5.5	30.0 ± 5.5	32.1 ± 6.1	32.1 ± 5.7	32.3 ± 5.9	31.3 ± 6.7	−0.44 (−1.41, 0.54)	0.38	0.88
RAVLT2	6.5 ± 2.5	6.3 ± 2.4	7.0 ± 2.2	7.2 ± 2.4	7.0 ± 2.5	7.0 ± 2.4	−0.15 (−0.53, 0.23)	0.44	0.63
RAVLT3	6.0 ± 2.6	5.6 ± 2.3	6.7 ± 2.6	6.6 ± 2.6	6.6 ± 2.6	6.7 ± 2.7	−0.19 (−0.58, 0.20)	0.33	0.09
RAVLT4	22.7 ± 1.5	22.6 ± 1.7	23.0 ± 1.5	23.0 ± 1.5	22.8 ± 1.4	22.8 ± 1.6	−0.18 (−0.44, 0.08)	0.18	0.10

¹Values are mean ± SD except TMT B, for which they are median [IQR]. A linear mixed-effects repeated-measures model was fitted to the data for each cognitive test. Test score was the dependent variable; randomized treatment, sex, and diabetic status were fixed effects; age and baseline score were covariates; subject was a random effect; and assessment time (12 or 30 mo) was a repeated measure. COWAT, Controlled Oral Word Association Test; DSST, Digit-Symbol Substitution Test; RAVLT, Rey Auditory Verbal Learning Test; RAVLT1, immediate recall as a sum of 4 trials; RAVLT2, short delay recall; RAVLT3, delayed recall; RAVLT4, delayed recognition; TMT, Trail Making Test.

²A lower score is better. For all other tests, a higher score is better.

To determine if the better outcome in those randomly assigned to EPA and DHA than control was due to an improvement with EPA and DHA or a decline in the control group, we compared results at 12 mo with baseline. As Table 4 shows, compared with baseline, improvement occurred at 12 mo in those randomly assigned to EPA and DHA in verbal fluency, language, and memory (COWAT 1, P = 0.047); psychomotor speed and attention (DSST, P < 0.001); and visual-motor coordination (TMT A and B, P = 0.033 and <0.001, respectively). In contrast, those randomly assigned to control had no change compared with baseline. Of note, memory and recall (RAVLT) improved in both those on EPA+DHA and the control group at 1 y compared with baseline, with the exception that RAVLT4 was improved at 1 y in the control group but not in the EPA+DHA group. However, there was no difference between the 2 groups for RAVLT4 over the 30-mo period, as Table 3 shows.

TABLE 4.

Within-group difference at 1 y compared with baseline for the EPA+DHA and control groups¹

	EPA+DHA			Control
Cognitive test	Mean difference	95% CI	P value	Mean difference	95% CI	P value
COWAT 1	0.8 ± 4.6	0.01, 1.6	0.047	0.08 ± 4.5	−0.7, 0.9	0.83
COWAT 2	−0.02 ± 5.2	−0.9, 0.9	0.95	−0.3 ± 4.9	−1.1, 0.6	0.51
DSST	2.7 ± 7.4	1.4, 3.9	<0.001	1.0 ± 9.3	−0.6, 2.6	0.23
TMT A2	−2.2 ± 11.8	−4.1, −0.2	0.033	0.6 ± 13.0	−1.7, 2.9	0.59
TMT B2	−2.4	—	<0.001	−1.0	—	0.48
RAVLT1	0.7 ± 4.8	−0.1, 1.5	0.068	2.0 ± 5.0	1.1, 2.9	<0.001
RAVLT2	0.5 ± 2.0	0.2, 0.8	0.005	0.8 ± 2.1	0.4, 1.1	<0.001
RAVLT3	0.7 ± 2.1	0.3, 1.0	<0.001	1.0 ± 1.9	0.6, 1.3	<0.001
RAVLT4	0.1 ± 1.5	−0.2, 0.4	0.44	0.4 ± 1.6	0.1, 0.7	0.006

¹Values are mean ± SD. Normally distributed variables were compared using paired t tests except for TMT B which was compared with Wilcoxon's signed-rank test owing to data being nonnormally distributed. Therefore, a median difference was calculated, but a 95% CI could not be generated. COWAT, Controlled Oral Word Association Test; DSST, Digit-Symbol Substitution Test; RAVLT, Rey Auditory Verbal Learning Test; RAVLT1, immediate recall as a sum of 4 trials; RAVLT2, short delay recall; RAVLT3, delayed recall; RAVLT4, delayed recognition; TMT, Trail Making Test.

²A lower score is better. For all other tests, a higher score is better.

Post hoc exploratory comparisons

Post hoc exploratory comparisons for age, sex, and diabetes status over the 30-mo period showed no difference for the EPA+DHA group compared with the control group.

Discussion

In the current trial, subjects randomly assigned to EPA and DHA had significantly better cognitive function scores for verbal fluency, language, and memory and 2 tests of visual-motor coordination than the control group over the 30-mo period. The better cognitive scores in those randomly assigned to high-dose EPA and DHA were due to improvement at 1 y in verbal fluency, language, and memory and 2 tests of visual-motor coordination compared with their baseline, whereas those randomly assigned to control had no change. These trends persisted over the 30-mo period. The improvement in the EPA and DHA group is clinically important and suggests that EPA and DHA should be considered to improve cognitive function. Although no difference was observed in outcome in subgroup analyses by age, sex, and diabetes status, it is possible that our sample size was too small. Taken together, our results are important because they suggest that high-dose EPA and DHA can result in sustained improvement over a 30-mo period in several cognitive tests in cognitively healthy older adults that could potentially prevent or delay cognitive decline in an aging population.

Table 5 summarizes prior trials which have examined the effect of EPA and DHA on cognitive function in cognitively healthy subjects aged >50 y. We conducted a systematic search using the terms omega-3 fatty acid, EPA, DHA, and cognitive function for original articles and conducted a systematic search for meta-analyses using the same search terms plus meta-analysis. Trial duration ranged from 1 mo to 6 mo for 7 trials and 2, 3.3, and 5 y for 3 trials with the total dose of EPA and DHA ranging from 285 mg to 2.55 g daily. Of these 10 trials, Witte et al. (28) reported a benefit on verbal fluency (COWAT) (P = 0.009) and a composite executive score (P = 0.023) over a 26-wk period with a dose of 1320 mg EPA plus 880 mg DHA per day compared with placebo. Moreover, the preservation of both the integrity of the microstructure of white mass and the volume of gray mass was observed. However, no significant change was observed in a composite score for memory (P = 0.06) or sensorimotor speed and attention (P > 0.26) (28). Külzow et al. (29) randomly assigned 44 subjects to 2.2 g EPA and DHA or sunflower oil placebo over 26 wk and reported significant improvement in the recall of object locations but no benefit on the more commonly used cognitive tests which we used (Table 5). Konagai et al. (30) randomly assigned 45 healthy males with a mean ± SD age of 67 ± 3.4 y to 12 wk of either medium-chain triglycerides as placebo, krill oil (consisting of 193 mg EPA and 92 mg DHA incorporated as phosphatidylcholine), or sardine oil (consisting of 491 mg EPA and 251 mg DHA incorporated in triglycerides). During performance of a working memory task (assessed by P300), a significant increase in cerebral cortex oxyhemoglobin concentration was noted in both the krill oil (P = 0.004) and sardine oil (P = 0.043) groups as compared with the placebo, a finding suggesting activation of cognitive function in older adults (30). However, they did not examine cognitive function with neuropsychological cognitive testing. Benefit was observed in a 1-mo study which assessed working memory using P300 testing (31). In summary, of the 10 studies, 2 showed benefit on cognitive function measured by neuropsychological testing at 6 mo. A third study of 1-mo duration showed a benefit; however, this is not long enough for assessment of long-term benefit on cognitive function. In contrast, the other 7 studies reported no benefit in cognitively healthy older adults (32–37).

TABLE 5.

Summary of randomized controlled trials of ω-3 fatty acids in cognitively healthy subjects¹

	Subjects, n; age range; mean ± SD age, y; study length; % male		P values
							RAVLT
Authors		Intervention	COWAT	DSST	TMT A	TMT B	RAVLT1	RAVLT2	RAVLT3	RAVLT4
Külzow et al.2 (29)	n = 44;50–75 y;Age: EPA/DHA: 63 ± 6, Placebo: 61 ± 6;26 wk;45% male	EPA 1320 mg, DHA 880 mg, both with vitamin E 15 mg;Placebo: sunflower oil 406 mg	Words starting with S: 0.25;Fruits: 0.62			0.55	0.33		0.48	0.87
Howe et al. (36)	n = 38;40–85 y;Age: EPA: 63.2 ± 1.6, Placebo: 64.1 ± 2.3;20 wk;68% Male;Mildly hypertensive	EPA 400 mg, DHA 1600 mg;Placebo: corn oil			P = 0.5 for the sum of TMT A and TMT B
van de Rest et al. (34)	n = 302;≥65 y;Age: EPA/DHA: 1800 mg: 69.9 ± 3.4, EPA/DHA: 400 mg: 69.5 ± 3.2, Placebo: 70.1 ± 4.0;26 wk;55% male	High dose:EPA 1093 mg, DHA 847 mg;Low dose:EPA 226 mg, DHA 174 mg;Placebo: high-oleic sunflower oil	NS		NS	NS	NS		NS	NS
Dangour et al.3 (35)	n = 867;70–79 y;Age: fish oil: 74.7 ± 2.5, Placebo: 74.6 ± 3.0;24 mo;55% male	EPA 200 mg, DHA 500 mg;Placebo: olive oil	NS	NS			0.14		0.46
Chew et al.4 (37)	n = 3501;50–85 y;Age: 72.7 ± 7.7;5 y;42.5% male	EPA 650 mg, DHA 350 mg;Vs. AREDS formulation: vitamins C, E, β-carotene, and zinc	Animals: 0.34;Letters starting with F, A, and S: 0.78
Witte et al.5 (28)	n = 65;50–75 y;Age: 63.9 ± 6.6;26 wk;53.9% male	EPA 1320 mg; DHA 880 mg;Placebo: sunflower oil	0.009
Mazereeuw et al. (32); CAD subjects	n = 92;Age: 61.7 ± 8.7;12 wk;76% male	DHA 1200 mg/d, EPA 600 mg/d, EPA:DHA ratio 1:2;Other: LC-PUFAs 100 mg/d;Placebo: soybean/corn oil	NS			NS	No benefit on CVLT which is similar to RAVLT
Tokuda et al. (31)	n = 113;Age: Placebo: 59.5 ± 0.4, EPA/DHA: 59.8 ± 0.5;1 mo;100% male	DHA 300 mg, EPA 100 mg,ARA 120 mg;Placebo: olive oil	Improvement in P3006
Geleijnse et al. (33); post-MI patients	n = 2911;Age: men: 68.7 ± 5.5 y, women: 70.3 ± 5.6 y;40 mo	1) EPA:DHA 400 mg/d, EPA:DHA ratio 3:2; 2) EPA/DHA 400 mg/d, ALA 2 g/d, EPA:DHA ratio 3:2;Placebo: margarine	No change in MMSE
Konagai et al. (30)	n = 45;Age: 67.1 ± 3.4 y;3 mo;100% male	1) Krill oil: DHA 92 mg/d, EPA 193 mg/d, DHA:EPA ratio 1:2;2) Sardine oil: DHA 251 mg/d, EPA 491 mg/d;Placebo: medium-chain triglyceride	Both krill oil and sardine oil had better P300 6 than placebo

¹ARA, arachidonic acid; AREDS, Age-Related Eye Disease Study; CAD, coronary artery disease; COWAT, Controlled Oral Word Association Test; CVLT, California Verbal Learning Test; DSST, Digit-Symbol Substitution Test; LC-PUFA, long-chain PUFA; MMSE, Mini-Mental Status Examination; NS, nonsignificant; RAVLT, Rey Auditory Verbal Learning Test; RAVLT1, immediate recall as a sum of 4 trials; RAVLT2, short delay recall; RAVLT3, delayed recall; RAVLT4, delayed recognition; TICS-M, Telephone Interview Cognitive Status-Modified; TMT, Trail Making Test.

²Külzow et al. (29) used the LOCATO, which examines visual-spatial object location of buildings, and the RAVLT. After adjustment for changes in diet, those on ω-3 fatty acids did better on the LOCATO (P = 0.004). As noted in the table, the RAVLT was not significant.

³The CVLT was performed which is similar to the RAVLT.

⁴This is an ancillary study of the AREDS2 randomized clinical trial. A composite score was computed to obtain an overall score for a battery of 8 cognitive function tests and was not significant. The composite score included the scores of the following: TICS-M word list for immediate recall; Verbal fluency with animal category; Verbal fluency with letter category; Verbal fluency alternating between animal and letter categories; Wechsler logical memory I and Wechsler logical memory II which measured immediate and delayed recall of 2 stories, respectively; Digits backward; and TICS-M delayed recall. All test results were converted into z scores which were then added.

⁵Witte et al. (28) calculated a composite executive score which was significant at P = 0.023. The composite executive score was defined as the sum of phonemic fluency and semantic fluency minus TMT (part B − part A)/part A and minus STROOP {[part 3 − (part 1 + part 2)]/2}/4. Preservation of both the integrity of the microstructure of white mass and the volume of gray mass was observed with a dose of 1320 mg EPA plus 880 mg DHA per day for 26 wk.

⁶P300 measures working memory by measuring event-related potentials using an electroencephalograph (31). P300 depicts central nervous system activity involved in the processing of new information when attention is engaged to update memory representation. P300 has 2 components: latency, which reflects cognitive processing speed; and amplitude, which measures attentional resources engaged in task completion (30). Latency is prolonged and amplitude is decreased with aging, a finding suggesting that P300 testing can evaluate cognitive function in healthy elderly adults (31).

A recent meta-analysis of cognitive function in 25 studies of both cognitively healthy and cognitively impaired subjects reported a mild benefit on memory function but no effect on global cognitive function measured by the MMSE (38). In another meta-analysis, Mazereeuw et al. (39) reported improvement in the immediate recall subcategory of episodic memory in nondemented adults. The hippocampus is important for memory registration and recall (40), and its volume decreases with age (41), a change which may be associated with cognitive dysfunction (40–42). Along these lines, fish oil supplementation reduced hippocampal atrophy in a cohort of healthy subjects and those with mild cognitive impairment (41). In a meta-analysis of cognitive function which included 38 randomized controlled trials in adults of ≥24 wk duration, Brainard et al. (43) reported a significant but clinically unimportant benefit on MMSE, differing by <1% of baseline. However, they found no effect on new neurocognitive outcomes or cognitive impairment in other cognitive domains. Zhang et al. (44) pooled data from 6 pre-2015 randomized controlled trials and found significant but probably clinically unimportant differences in MMSE, a finding similar to that of Brainard et al.’s meta-analysis (43).

In contrast to these trials, our trial used a higher dose, 3.36 g daily, for a longer period of time: 30 mo. Therefore, to our knowledge, our trial is the first to show a long-term benefit of high-dose EPA and DHA on multiple cognitive tests in cognitively healthy older adults. Of note, the placebo in some of the prior studies was high-oleic acid sunflower oil or olive oil, both of which are anti-inflammatory, raising the issue of whether they may dilute benefit of the ω-3 fatty acid examined and therefore not be optimal placebos (see Table 5) (28, 29, 31, 34, 35).

The current findings of an improvement in memory at 1 y (measured by RAVLT) in both the EPA+DHA and control groups are of interest. We used the same word list at baseline, 1 y, and 30 mo and there may have been a “learning effect.” The other tests, COWAT 1 and TMT A and B, which were improved in the ω-3 group but not in the control group, test visual-motor coordination (TMT A and B) and verbal fluency (COWAT 1) and are not subject to a learning effect. Therefore, the learning effect we observed with RAVLT suggests that it is important to use different word lists with sequential testing. This may be 1 reason for lack of benefit being observed in other studies using RAVLT and its versions.

The current findings with verbal fluency (COWAT) are of special interest. COWAT requires subjects to name all the words they can think of for 2 separate lists. Our lists were animals (COWAT 1) and supermarket items (COWAT 2). Significant benefit was noted for the animal list, but no significant benefit was noted for the list of supermarket items. This finding suggests that the list should be carefully selected to ensure optimal assessment of cognitive function and that supermarket items may not be as readily recognized, especially by male subjects, as more common items such as animals.

Strengths and limitations

Major strengths of the present study are the randomized controlled design, the use of high-dose (3.36 g) EPA and DHA, the large sample size, the completeness of baseline data and of cognitive testing, and the long treatment duration of 30 mo. A limitation is that our subjects had clinical CAD which limits generalizability; however, the age range is that in which older subjects are developing memory problems; thus, consideration should be given for ω-3 fatty acids to improve cognitive function. Although our trial was open-label, cognitive testing is objective; therefore, the open-label nature should not have influenced the results.

Conclusions

The results of the present study in older, cognitively healthy adults with clinical CAD show that those randomly assigned to high-dose EPA and DHA had better cognitive function at 1 y than at baseline and significantly better cognitive function over 30 mo than control subjects. These findings have potential public health significance and should encourage consideration of ω-3 fatty acids in cognitively healthy subjects to improve cognitive function that could potentially prevent or delay cognitive decline in an aging population. These findings are especially important for patients with CAD because CAD has been implicated as a risk factor for dementia.

Notes

The data in this article were presented at the American Heart Association Scientific Sessions, Philadelphia, PA, 17 November 2019.

Supported by NIH Specialized Center of Clinically Oriented Research program grant P50 HL083813 (to FKW) and Harvard Clinical and Translational Science Center award NIH UL1 TR001102 (to FKW). GlaxoSmithKline (Research Triangle Park, NC) provided Lovaza.

Supplemental Figure 1 is available from the “Supplementary data” link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/ajcn/.

AM, AR, and BV contributed equally to this work.

Abbreviations used: CAD, coronary artery disease; COWAT, Controlled Oral Word Association Test; DSST, Digit-Symbol Substitution Test; MMSE, Mini-Mental Status Examination; RAVLT, Rey Auditory Verbal Learning Test; TMT, Trail Making Test.

Data Availability

Data will be made available upon request pending application and approval.