Fish oil supplementation and risk of dementia among diabetic patients: a prospective study of 16,061 older patients

Abstract

Background

Although n-3 Polyunsaturated fatty acids (PUFAs) may benefit cognitive performance, the association of n-3 PUFA intake with dementia risk under dysglycemia has not been examined. We aimed to evaluate the relationship between fish oil supplement use or fish consumption and dementia risk among older patients with diabetes.

Method

A total of 16,061 diabetic patients aged over 60 years were followed up in the UK Biobank. Fish oil supplements use (yes or no) was collected by the touch screen questionnaire. The diagnosis of dementia was ascertained by the UK Biobank Outcome Adjudication Group. The hazard ratios (HRs) and 95% confidence intervals (95% CIs) were estimated using Cox proportional hazards models.

Results

A total of 337 cases of dementia were confirmed after a mean duration of 7.7 years (123,486 person-years) of follow-up. Habitual use of fish oil supplements showed a 24% lower dementia risk among older diabetic patients [HRs (95% CIs): 0.76 (0.60−0.98) (P = 0.031)] compared with non-users. Such inverse association was not modified by the APOE ε4 genotype. However, the consumption of both oily fish (≥2 times/week) and non-oily fish (≥2 times/week) had no significant association with dementia risk (p-trend = 0.271 and p-trend = 0.065) compared with non-consumers.

Conclusion

In summary, fish oil supplementation may play a protective role in cognitive function across all APOE genotypes, while non-oily fish and oily fish consumption have no protective association among older diabetic patients.

1. Introduction

The number of diabetic patients is continuously climbing, and up to 537 million diabetic patients have been diagnosed all around the world [1]. Diabetes could impair brain function and diabetic patients have a doubled risk of dementia compared to people without diabetes, which would lead to serious implications in nations with increasing incidences of diabetes [2]. The possible mechanism of diabetes in accelerating cognitive decline included cerebral insulin resistance, vascular endothelial dysfunction resulting from the accumulation of lipids, aggregated protein and advanced glycation end products (AGEs), inflammation, peroxidative membrane injury, and mitochondrial dysfunction [3]. A broad concern should be paid to older diabetic patients with a high risk of dementia for the prevention of cognitive impairment.

Marine n-3 polyunsaturated fatty acid (PUFA) supplementation could play a beneficial role in the promotion of cerebral health and the reduction of incidental diseases, including Alzheimer’s disease (AD) and diabetes [4]. A recent meta-analysis revealed that higher dietary n-3 PUFA intake was related to lower risk of mild cognitive impairment (MCI) [5]. Two studies conducted in the UK Biobank study found fish oil supplementation could prevent dementia among older adults [6,7]. But, fish consumption, abundant in n-3 PUFA, was not related to dementia risk in a recent meta-analysis [8]. In addition, a meta-analysis of randomized control trials (RCTs) suggested no effect of n-3 PUFA supplementation on cognitive function [9]. The effect of n-3 PUFA supplementation on cognitive function in healthy individuals remains uncertain. Importantly, n-3 PUFA supplementation from diet or supplements in relation to incident dementia among diabetic patients is still unknown.

Regarding diabetes, some evidence indicated that n-3 PUFA intake could improve the cognitive function of diabetic rats or mice by various pathways. The related mechanism included modulating insulin resistance [10], improving mitochondrial dysfunction [11], reducing neuroinflammation [12], and inducing abnormal expression of brain-derived neurotrophic factor (BDNF) [13]. However, evidence from epidemiological studies among diabetic patients is still absent.

To fill this gap, we performed a large cohort study to examine the association between n-3 PUFA intake and cognitive function among diabetic patients (> 60 years old) from the UK Biobank.

2. Materials and methods

2.1. Study population

UK Biobank involved approximately 0.5 million participants aged 40–73 years. They were recruited from 22 assessment centers in the UK during 2006−2010. Demographic, lifestyle, diet, and health-related information, including the history of disease and medication use, were collected from a touch screen questionnaire or face-to-face interview. Data about physical measurement and biological samples were also acquired in the assessment center. Overall, 28,033 diabetic patients engaged in the UK Biobank. After excluding the patients who were aged < 60 years old, with dementia at baseline, and, or lack of data about fish oil use, 16,061 older patients with diabetes were finally included in our analysis (Supplementary Fig. S2).

2.2. Assessment of fish oil use, fish consumption, and n-3 PUFA index

Participants needed to answer some questions about diet and health state by a touch screen questionnaire. For example, “Do you use the following supplements regularly?” They had two options listed below the question to answer fish oil use (yes or no). Moreover, the two repeated surveys had successively been conducted during 2012–2013 and in 2014 later to assess the reproducibility of fish oil use. Approximately 78% of fish oil users insisted on supplementing fish oil in the first repeated survey (Spearman r = 0.70), and 55% of fish oil users took the supplements in the second repeated survey (Spearman r = 0.44) (Supplementary Fig. S3).

A food frequency questionnaire (FFQ) involved 29 questions about diet in the past year and was used to collect dietary information in the baseline survey. The frequencies of taking both non-oily fish and oily fish included ≥two times/week, one time/week, <one time/week, and never.

Plasma fatty acid level was measured by magnetic resonance (NMR) at baseline, including n-3 PUFA, docosahexaenoic acid (DHA), n-6 PUFA, saturated fatty acids (SFA), and monounsaturated fatty acid (MUFA). The definition of n-3 PUFA index, DHA index, or other n-3 PUFA index was expressed as a percentage of total fatty acid in plasma [14].

2.3. Covariates

The touch screen questionnaire involved demographic and lifestyle information, including age, race, sex, living environment, education levels, alcohol consumption, physical activity, and smoking. Drug use and the history of disease were also collected by the touch screen questionnaire. Body mass index (BMI) was defined as weight (kg) divided by height squared (m²). Education levels were classified into several categories, including national exams, vocational qualifications, college or university degrees, and others. An indicator of material deprivation was calculated, and Townsend deprivation index (TDI) incorporated non-car ownership, unemployment, household overcrowding, and non-home ownership [15]. International Physical Activity Questionnaire was employed to reckon the metabolic equivalent of task (MET), and physical activity level was assessed by MET hours per week [16]. A total of ten types of foods, including fruits, vegetables, refined grains, vegetable oils, whole grains, fish, unprocessed meat, processed meat, sugar-sweetened beverages, and dairy, were used to compute a healthy diet score. The ideal food intake was based on the definition of diet for cardiometabolic health (Supplementary Table S1) [17,18]. Individuals would get 1 point when they had achieved the goal of a specific food intake. In total, the diet score was 10. The higher the score was, the healthier diet the person had. The healthy lifestyle score was evaluated according to a healthy diet score ≥5, physical activity ≥600 MET-h/week, BMI < 30 kg/m², and non-smoking status [19].

The frequencies of other food groups’ consumption ranged from “never” to “once or more daily”. Considering the influence of the duration of diabetes among patients, we separated diabetic participants into four groups, representing the duration of diabetes, referring to ≥10 years, 5−10 years, 0–5 years, or missing [20].

Blood samples were collected by qualified physicians at baseline and measured to evaluate biochemical indexes, including C-reactive protein (CRP), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), total cholesterol (TC), triglyceride (TG), and apolipoprotein B (ApoB), and apolipoprotein A (ApoA). Given a right-skewed distribution of serum CRP level, converted CRP values (log (units+1)) were employed in our analysis [21]. More details can be achieved from the website (https://biobank.ctsu.ox.ac.uk/showcase).

2.4. APOE genes

Genotyping was conducted by the UK BiLEVE arrays Axiom (UK Biobank Axiom Array, ThermoFisher). Both rs429358 and rs7412 determined the phenotype of APOE ε4 status according to the genome-wide genotyping [22]. The phenotypes of ε4/ε4 or ε3/ε4 represented a high risk of dementia, ε2/ε3 or ε2/ε2 for a low risk, and ε3/ε3 or ε2/ε4 for a normal risk [23].

2.5. Ascertainment of incident dementia

The definition of dementia, including Alzheimer’s disease, vascular dementia, frontotemporal dementia, and other dementia, was identified by the UK Biobank Outcome Adjudication Group, and the data concerning dementia was achieved from hospital inpatient records, self-report, and death registries. Incident dementia was recorded using ICD-10 codes, or ICD-9 codes, and the codes are shown in Supplementary Table S2. The shorter interval between the initial evaluation and the dementia diagnosis, death, the last follow-up (March 31, 2017), or loss to follow-up was calculated as the follow-up duration.

2.6. Statistical analysis

The Cox proportional hazards model was applied to evaluate the role of fish oil supplementation on cognitive function among older diabetic patients. We adjusted for multiple variables to control potential confounders. Model 1 was adjusted for age, sex, BMI, smoking status, physical activity, alcohol consumption, race, TDI, education level, assessment center, history of hypertension, history of cardiovascular disease, APOE ε4 status, history of depression, and duration of diabetes. We further adjusted for the consumption of oily fish, non-oily fish, fruits, vegetables, processed red meat, and whole grains in model 2. For model 3, a healthy diet score was further adjusted based on model 1.

Subgroup analyses were conducted to detect the potential interaction between fish oil supplementation and other confounding factors by the likelihood ratio test. Adding product terms of fish oil use and stratifying variables into the model was used to estimate the interaction. Stratifying variables included age, sex, BMI, smoking status, physical activity, alcohol consumption, TDI, history of hypertension, history of cardiovascular disease, APOE ε4 status, oily fish intake, non-oily fish intake, history of depression, the duration of diabetes, healthy lifestyle score, and healthy diet score.

To test the robustness of the findings, we performed several sensitivity analyses. Use of medication (lipid-lowering drugs, hypoglycemic drugs, or aspirin), other supplement use (vitamins or minerals), a history of hypercholesterolemia, or excluded participants with missing covariate data were additionally adjusted in the sensitivity analyses. Excluding incident dementia cases that emerged in the first two years of follow-up was also analyzed in the sensitivity analyses. It would be seen as significant when the P-value on the 2-side was <0.05. SAS version 9.4 was applied to conduct the above analyses (SAS Institute, Cary, NC, USA).

3. Results

3.1. Population characteristics

We divided the participants into two groups according to the categorical features of the users or non-users of fish oil supplements. The baseline characteristics of participants are shown in Table 1. Among enrolled 16,061 participants, 5,268 individuals supplemented fish oils. Fish oil users tended to be female and older, and had lower BMI, lower TDI, and higher educational levels than non-users. Regarding lifestyle, a smaller proportion of fish oil users were current smokers. More fish oil users had heavier physical activity and did not have histories of cardiovascular disease and depression, but they had a shorter duration time of diabetes. They also had a partiality towards consuming fish, fruits, whole grains, and vegetables, whereas they were unlikely to consume refined grains, coffee, sugar drinks, and processed red meat. Generally, fish oil users had a healthier diet than fish oil non-users.

Table 1.

Basic characteristics of diabetic patients aged >60 years by the use of fish oil supplements in the UK Biobank cohort.

	Overall	Fish oil non-users	Fish oil users	Pa
Characteristics	(n = 16,061)	(n = 10,793)	(n = 5,268)
Male, %	63.2	64.1	61.4	<0.001
Age, yrb	64.6 (2.9)	64.5 (2.9)	64.7 (2.9)	<0.001
Race, %				<0.001
White	90.7	90.6	90.8
Asian	4.9	5.4	3.9
Black	2.3	1.9	3.1
Mixed	0.5	0.4	0.6
Others	1.2	1.2	1.2
BMI, kg/m2b	30.9 (5.4)	31.0 (5.5)	30.6 (5.2)	<0.001
Townsend deprivation indexb	−0.7 (3.3)	−0.6 (3.3)	−1.0 (3.2)	<0.001
Education, %				0.009
College or University degree	20.4	20.1	21.1
Vocational qualifications	15.2	15.0	15.6
Optional national exams at ages 17−18 years	8.1	7.7	8.8
National exams at age 16 years	20.7	20.7	20.8
Others	33.7	34.5	32.0
Physical activity, MET-h/wkb	37.8 (42.0)	35.8(41.2)	41.8 (43.4)	<0.001
Smoking status, %				<0.001
Never	40.8	40.7	41.0
Previous	49.5	48.8	50.8
Current	8.9	9.7	7.2
Alcohol consumption, %				0.001
Never	15.2	15.8	14.1
Special occasions only	17.9	18.2	17.2
1−3 times/month	11.0	11.1	10.8
1 or 2 times/week	22.9	22.4	23.8
3 or 4 times/week	16.1	15.4	17.5
Daily or almost daily	16.8	16.9	16.4
APOE ε4 status, %
ε2/ε2 or ε2/ε3	12.8	12.5	13.5
ε3/ε3 or ε2/ε4	58.2	58.2	58.1
ε3/ε4 or ε4/ε4	23.8	23.9	23.6
History of disease, %
Hypertension	88.8	88.7	88.8	0.928
Cardiovascular disease	14.8	15.5	13.3	<0.001
Depression	20.9	21.5	19.6	<0.001
Duration of diabetes, %				<0.001
0−5 years	42.9	42.2	44.3
5−10 years	27.4	27.0	28.0
≥10 years	29.4	30.4	27.4
Dietary consumption, %
Oily fish, times/week				<0.001
<1.0	39.1	42.2	32.9
1.0	38.2	37.1	40.5
≥2.0	21.9	19.9	26.0
Non-oily fish, times/week				<0.001
<1.0	30.1	31.5	27.4
1.0	51.3	50.6	52.7
≥2.0	17.7	17.0	19.2
Poultry, times/week				0.299
<2.0	54.1	54.7	53.1
2.0−4.0	43.2	42.7	44.2
>4.0	2.2	2.2	2.3
Processed meat, times/week				0.002
<1.0	32.8	32.0	34.5
1.0	30.0	29.9	30.2
≥2.0	36.9	37.7	35.1
Unprocessed red meat, times/week				0.033
<2.0	42.2	42.0	42.5
2.0−4.0	46.2	45.8	46.9
>4.0	11.5	12.0	10.5
Vegetables, servings/day				<0.001
<1.0	17.2	18.6	14.3
1.0−2.9	72.1	71.2	73.8
≥3.0	9.8	9.2	11.2
Fruits, servings/day				<0.001
<2.0	27.5	29.4	23.5
2.0−3.9	49.3	48.8	50.5
≥4.0	22.9	21.6	25.7
Whole grains, servings/day				<0.001
<1.0	40.6	43.5	34.7
1.0−2.9	41.6	39.7	45.6
≥3.0	17.1	16.1	19.2
Refined grains, servings/day				<0.001
<1.0	51.9	49.1	57.7
1.0−2.9	34.4	35.8	31.6
≥3.0	13.0	14.4	10.1
Cheese, times/week				0.953
<2.0	47.0	47.0	47.0
2.0−4.0	39.4	39.3	39.5
>4.0	9.0	9.0	8.9
Coffee, cups/day				<0.001
<1.0	29.4	29.9	28.2
1.0−2.0	37.6	35.8	41.3
≥3.0	32.7	33.9	30.2
Sugar-sweetened beverages consumer, %	48.5	49.1	47.3	0.034
Healthy diet scoreb	3.4 (1.5)	3.3 (1.5)	3.6 (1.5)	<0.001

^aχ² test was used to calculate P values and t test for continuous variables.

^bValues are means ± SD or percentages unless stated otherwise.

3.2. The association of fish oil supplementation, fish intake, and n-3 PUFA index with incidental dementia among older diabetic patients

As shown in Fig. 1, 337 incident cases of dementia were identified among 16,061 individuals after a mean duration of 7.7 years of follow-up. Fish oil supplementation was significantly related to incidental dementia among older diabetic patients [HRs (95% CIs): 0.69 (0.54−0.89), P =  0.003] in age- and sex-adjusted model. Compared with non-users, the HRs (95% CIs) was 0.76 (0.59−0.97) after adjustment for demographic, and lifestyle variables in model 1. After additional adjustments for five important dietary factors (oily fish, vegetables, fruits, whole grains, processed red meat) based on model 1, fish oil use still had a protective role among older diabetic patients. Considering the influence of dietary patterns, a healthy diet score was additionally adjusted, and the protective association persisted.

Fig. 1.

Association between fish oil supplementation and incident dementia among diabetic patients aged >60 years in the UK Biobank (n = 16,061).

Cox proportional hazards models was used to get hazard ratios (HRs) and 95% confidence intervals (95% CIs). Model 1 was adjusted for age, sex, race (white, Asian, black, mixed, or other ethnic group), location of assessment centers (22 categories), BMI (in kg/m²; <18.5, 18.5–25, 25–30, 30–35, ≥35, or missing), education (college or university degree, vocational qualifications, optional national exams at ages 17−18 years, national exams at age 16 years, others, or missing), TDI (tertiles), smoking status (never, former, current, or missing), alcohol consumption (never, special occasions only, 1–3 times/month, 1 or 2 times/week, 3 or 4 times/week, or daily/almost daily), physical activity (in MET-h/wk; tertiles), history of hypertension (yes or no), history of cardiovascular disease (yes or no), APOE ε4 status (ε2/ε2 or ε2/ε3, ε3/ε3 or ε2/ε4, ε3/ε4 or ε4/ε4), history of depression (yes or no), duration of diabetes (0−5 years, 5−10 years, or ≥10 years). Model 2 was further adjusted for oily fish (<1, 1, or ≥2 times/week), non-oily fish (<1, 1, or ≥2 times/week), vegetables (<1, 1–2.9, or ≥3 servings/day), fruits (<2.0, 2.0−3.9, or ≥4.0 servings/day), whole grains (<1.0, 1.0−2.9, or ≥3.0 servings/day), processed red meat (<1, 1, or ≥2 times/week) based on model 1. Model 3 was further adjusted for healthy diet score (tertiles) based on model 1.

Both the consumption of non-oily fish and oily fish did not have protective associations with dementia risk among older diabetic patients (Fig. 2, Fig. 3). The HRs (95% CIs) showed 0.70 (0.46-1.08) and 1.13 (0.65–1.94) for oily fish intake (p-trend = 0.271) and non-oily fish intake (p-trend = 0.065), respectively.

Fig. 2.

Association between oily fish consumption and incident dementia among diabetic patients aged >60 years.

Cox proportional hazards models was used to estimate hazard ratios (HRs) and 95% confidence intervals (95% CIs). Model 1 was adjusted for age, sex, race (white, Asian, black, mixed, or other ethnic group), location of assessment centers (22 categories), BMI (in kg/m²; <18.5, 18.5–25, 25–30, 30–35, ≥35, or missing), education (college or university degree, vocational qualifications, optional national exams at ages 17−18 years, national exams at age 16 years, others, or missing), TDI (tertiles), smoking status (never, former, current, or missing), alcohol consumption (never, special occasions only, 1–3 times/month, 1 or 2 times/week, 3 or 4 times/week, or daily/almost daily), physical activity (in MET-h/wk; tertiles), APOE status (ε2/ε2 or ε2/ε3, ε3/ε3 or ε2/ε4, ε3/ε4 or ε4/ε4). Model 2 was additionally adjusted for history of hypertension (yes or no), history of cardiovascular disease (yes or no), history of depression (yes or no), duration of diabetes (0−5 years, 5−10 years, or ≥10 years) based on model 1. Model 3 was further adjusted for fish oil supplementation (yes or no), vegetables (<1, 1–2.9, or ≥3 servings/day), fruits (<2.0, 2.0−3.9, or ≥4.0 servings/day), whole grains (<1.0, 1.0−2.9, or ≥3.0 servings/day), processed red meat (<1, 1, or ≥2 times/week), unprocessed red meat (<1, 1, or ≥2 times/week), poultry (<2, 2–4, or ≥4 times/day), non-oily fish (never, <1, 1, or ≥2 times/week) and coffee (<1, 1–2, or ≥3 cups/day).

Fig. 3.

Association between non-oily fish consumption and incident dementia among diabetic patients aged >60 years.

Cox proportional hazards models was used to estimate hazard ratios (HRs) and 95% confidence intervals (95% CIs). Model 1 was adjusted for age, sex, race (white, Asian, black, mixed, or other ethnic group), location of assessment centers (22 categories), BMI (in kg/m²; <18.5, 18.5–25, 25–30, 30–35, ≥35, or missing), education (college or university degree, vocational qualifications, optional national exams at ages 17−18 years, national exams at age 16 years, others, or missing), TDI (tertiles), smoking status (never, former, current, or missing), alcohol consumption (never, special occasions only, 1–3 times/month, 1 or 2 times/week, 3 or 4 times/week, or daily/almost daily), physical activity (in MET-h/wk; tertiles), APOE status (ε2/ε2 or ε2/ε3, ε3/ε3 or ε2/ε4, ε3/ε4 or ε4/ε4). Model 2 was additionally adjusted for history of hypertension (yes or no), history of cardiovascular disease (yes or no), history of depression (yes or no), duration of diabetes (0−5 years, 5−10 years, or ≥10 years) based on model 1. Model 3 was further adjusted for fish oil supplementation (yes or no), vegetables (<1, 1–2.9, or ≥3 servings/day), fruits (<2.0, 2.0−3.9, or ≥4.0 servings/day), whole grains (<1.0, 1.0−2.9, or ≥3.0 servings/day), processed red meat (<1, 1, or ≥2 times/week), unprocessed red meat (<1, 1, or ≥2 times/week), poultry (<2, 2–4, or ≥4 times/day), non-oily fish (never, <1, 1, or ≥2 times/week) and coffee (<1, 1–2, or ≥3 cups/day).

No significant association between n-3 PUFA index, DHA index, or other n-3 PUFA index and incidental dementia was found in multivariable-adjusted models (Supplementary Table S3). The HRs (95% CIs) were 0.75 (0.36–1.56), 0.75 (0.31–1.83), and 1.01 (0.44–2.31) for n-3 PUFA index (p-trend = 0.55), DHA index (p-trend = 0.62), and other n-3 PUFA index (p-trend = 0.96), respectively.

3.3. Subgroup analyses

In subgroup analyses, the relationship was not modified by age, TDI, physical activity, alcohol consumption, smoking status, BMI, oily fish intake, Apolipoprotein E (APOE) ε4 status, history of hypertension, history of cardiovascular disease, history of depression, duration of diabetes, healthy lifestyle score, and healthy diet score (Fig. 4). A significant inverse relationship between fish oil supplementation and the risk of dementia was observed for individuals consuming non-oily fish ≥ one time/week. In contrast, no significant association was found among participants consuming non-oily fish <1 time/week.

Fig. 4.

The association of fish oil supplementation with the risk of dementia in subgroups.

BMI = body mass index; MET = metabolic equivalent of task. Forest plots show the multivariable HRs of dementia associated with fish oil use in subgroups. HRs were adjusted for age, sex, race, assessment centers, BMI, education, TDI, smoking, alcohol consumption, physical activity, history of hypertension, history of cardiovascular disease, APOE ε4 status, mineral supplementation, history of depression, duration of diabetes, and healthy diet score.

3.4. Sensitivity analyses

In sensitivity analyses, the significant association remained unchanged after adjustment for lipid-lowering drug use, hypoglycemic drug use, aspirin use, or hypercholesterolemia at baseline, or excluding the participants with missing covariate data (Supplementary Fig. S1). Fish oil supplement use had a marginally significant association with the risk of dementia after further adjustment for other supplements such as vitamins and minerals (P =  0.053). The finding still existed after excluding the dementia cases that emerged during the first two years of follow-up (P =  0.050).

3.5. The association of fish oil supplementation with serum biomarkers

Fish oil supplementation was found to have a positive linear association with HDL-C level (p =  0.008) and ApoA level (p < 0.001). Still, it was not related to CRP, LDL-C, TC, TG, and ApoB (Supplementary Table S4).

4. Discussion

In this cohort study of older diabetic patients, an inverse association of fish oil supplementation with dementia risk was observed. However, we did not find a remarkable association between oily or non-oily fish consumption and dementia risk among these patients. Our findings provide a preventive strategy to improve cognitive function for older diabetic patients.

4.1. Comparison with other studies

Expectedly, fish oil supplement use could decrease dementia risk among older diabetic patients. Consistent with our findings, some epidemiological studies have uncovered that supplementing n-3 PUFAs had a beneficial role in cognitive performance among older people. Following up 215,083 general older people up to 7.92 years, fish oil supplementation was inversely related to dementia risk in the UK Biobank study [HR (95% CI): 0.87 (0.79−0.96); P = 0.004] [7]. A similar finding was found in the UK Biobank study involving 211,094 older persons [6]. Moreover, a larger beneficial role of fish oil supplementation for diabetic patients than healthy people was observed, which may result from the protective role of fish oil supplement use on cognitive function through various pathways, including indirect and direct protection. Supplementation with n-3 PUFA could improve insulin resistance through the regulation of endoplasmic reticulum stress and mitochondria [24], which could avoid nerve damage from insulin resistance [25]. In addition, n-3 PUFA supplementation played a direct protection for synapses of nerves [26]. Older adults who were randomly assigned to 1 g fish oil supplement, including 430 mg DHA and 90 mg eicosapentaenoic acid (EPA) for 24 months, were found to have a better working memory than the individuals in the control group who received sunflower oil [27]. Moreover, a meta-analysis involving 25 RCT studies asserted that n-3 PUFA intake could mildly improve memory functions in non-demented older people (Hedge's g = 0.31, z = 2.945, P =  0.003) [28]. A meta-analysis involving 20 prospective studies found an inverse relationship between circulating n-3 PUFA and dementia risk [29]. Besides, supplementation with n-3 PUFA had a beneficial effect on various chronic diseases, including diabetes and cardiovascular disease (CVD). Higher consumption of marine n-3 PUFA was related to lower CVD risk in a meta-analysis including 25 cohort studies [30]. Supplementation with n-3 PUFA might have an indirect protective effect on cognitive function by directly improving glycolipid metabolism [10].

The promising effect of n-3 PUFA supplement on cognitive function was not found after supplementing 400 mg n-3 PUFA per day up to 12 weeks, and a reduction in TG was caught among patients with diabetes in this RCT study [31]. A total of 11,685 patients with dysglycemia were intervened with n-3 PUFA (1 g) or titrated basal insulin glargine or standard care or placebo in the Outcome Reduction with Initial Glargine Intervention (ORIGIN) trial, and no effect of n-3 PUFA supplementation on cognitive function was found [32]. No association was found in previous studies, which may not be enough time or the dose of n-3 PUFA supplementation.

Here we could not observe any inverse association of fish consumption with incidental dementia among diabetic patients. A recent meta-analysis found no relationship between fish consumption and dementia risk [RR (95% CI): 0.90 (0.79–1.02), n = 5] [8]. A positive relationship between a dietary pattern, emphasizing higher consumption of fish and vegetables, and MMSE score was found among diabetic patients in a cross-sectional study [33]. A cohort study conducted among 1,127 healthy participants showed that a high intake of fish might aid in preventing dementia in the Japan Public Health Center (JPHC)-based Prospective Study [OR (95% CI): 0.39 (0.18−0.86), P = 0.01] [34]. Although no association between both non-oily fish consumption and oily fish consumption and dementia risk was found, lower dementia risk was observed among fish oil users with higher non-oily fish consumption (≥one time/week). These results indicated that consuming non-oily fish seemed unable to reach an adequate intake level of n-3 PUFAs, but might be related to lower dementia risk in conjunction with fish oil supplement use.

No significant relationship between n-3 PUFA index and incidental dementia was found in our study, while a meta-analysis revealed a significant association between n-3 PUFA index and cognitive performance [8]. Besides, a cohort study involving the data on plasma fatty acids in 99,974 people found that individuals with higher plasma n-3 PUFA proportions had lower dementia risk [HR (95% CI) for per SD increment: 0.95 (0.91–1.00), P =  0.048] [35]. Only 3,838 older patients with diabetes were measured for plasma fatty acids and a total of 72 cases of dementia were diagnosed in our study. The small sample size with insufficient statistical power might explain the non-significant association between n-3 PUFA index and incidental dementia in our study. Further larger cohort studies should be conducted to investigate the relationship between n-3 PUFA index and cognitive function among diabetic patients.

Individuals with the APOE ε4 allele were at risk of developing AD [36], and the APOE ε4 allele might modify the relationship between n-3 PUFA consumption and cognitive functions [37]. Increasing APOE ε4 dosage could attenuate the protective role of fish oil supplements after following up with 445,961 participants from the UK biobank [38]. Dietary n-3 PUFA or the use of supplements could play a protective role in brain functions only for APOE ε4 noncarriers [39,40]. Nevertheless, fish oil supplementation was not modified by APOE genotypes in our analysis, which indicated the beneficial role was not different among individuals with different APOE genotypes. Shinohara et al. found that diabetes could damage cognitive function by inducing vascular impairment in APOE ε4 noncarriers, but the impairment was negligible to APOE ε4 carriers [41]. We observed similar effects of n-3 PUFA supplementation among diabetic patients with or without the APOE ε4, which indicated that n-3 PUFA supplementation could affect cognitive function through pathways independent of APOE ε4. A significant association between n-3 PUFA supplementation and dementia risk became marginally significant after further adjustment for other supplements. Other supplements use, such as vitamin or mineral supplements, could improve cognitive function [42,43], which might weaken the effect of n-3 PUFA supplementation.

Few studies paid attention to preventing cognitive impairment in older patients with diabetes with a high risk of dementia in the past. We found a beneficial effect of n-3 PUFA supplementation on cognitive function for older patients with diabetes in a longitudinal analysis. More large studies are needed to collaborate with our findings.

4.2. Potential explanations

As we all know, the risk of dementia increases with the aging process among healthy people, which is linked to aging-related brain function degradation. The pathological mechanism of dementia appears complex and is related to the change in cerebral fatty acids. Hosseini et al. discovered low levels of blood DHA in both MCI and AD patients, which might drive the occurrence of cognitive impairment [44]. Supplementation with n-3 PUFA attenuated cognitive damage from diabetes by increasing insulin sensitivity, anti-inflammatory, antiplatelet aggregatory, and antioxidant properties [4]. We summarized some possible mechanisms for elucidating the protective role of n-3 PUFAs on cognitive functions. It was regarded as an essential component in neuronal membranes and played a crucial role in transferring neuronal information [45]. A suitable proportion of n-6 and n-3 PUFAs could protect insulin receptor functions in cerebral cell membranes by maintaining membrane fluidity, which avoids changing downstream insulin receptors and damaging synaptic plasticity and cognition [10]. DHA could promote the formation of neuroprotectin D1 (NPD1) to suppress neurotoxicity [46] and directly protect against the apoptosis of neurons exposed to Aβ [47]. Besides, n-3 PUFAs could promote the formation of some anti-inflammatory mediators and inhibit pro-inflammatory cytokine production [48]. Moreover, n-3 PUFAs could protect vascular function by exerting antioxidant activity [49] and restore mitochondrial respiration by enhancing respiration control rate (RCR) and improving adenosine triphosphate levels [50]. In addition to the classic neuroprotective pathways above, n-3 PUFAs also could protect nerves against damage by the gut-brain axis and regulate cerebral carbohydrate metabolism. The expression of glucose transporter 1 (GLUT1) could be enhanced to accelerate the transportation of glucose into the brain by supplementing n-3 PUFA [51]. Supplementation with n-3 PUFA could interact with the pathogenesis of neurodegenerative disease by enhancing the amount of short-chain fatty acids (SCFAs), which would be beneficial to keeping the balance of gut microbiota [52]. Together, cognitive functions could benefit from n-3 PUFA intake. We advanced the recognition that n-3 PUFA supplement use should also be beneficial to lower the risk of dementia among diabetic patients, which links the association to carbohydrate metabolism. The causal relationship is still warranted to verify in the case of glucose homeostasis disorder.

4.3. Strengths

Several advantages of our study should be noted. First, the relationship between fish oil supplementation and dementia risk was innovatively demonstrated among older diabetic patients, which could fill the gap that previous studies highlighted prevention strategies for the general population rather than a high-risk population with glucose homeostasis disorder. Second, taking important confounding factors into account, including duration of diabetes and APOE genotypes, could attenuate bias and investigate the current association. Finally, a higher accuracy of the diagnosis existed in our analysis, which was proved by neuroimaging and laboratory reports in a validation study among 17,000 Biobank participants [53].

4.4. Limitations

Weaknesses also existed in our study. First, given a lack of data on long-term fish oil supplementation, we could not find a long-term association of habitual fish oil supplement use with incidental dementia among diabetic patients. Nonetheless, the validity and reproducibility of fish oil supplementation were assessed by two repeated surveys. Most fish oil users continued to use fish oil supplements at baseline. Second, unknown dosage and ratios of EPA to DHA were one of the main limitations, which might restrict us from formulating detailed guidance on preventing dementia among high-risk individuals. Third, we acknowledged bias related to residual confounding that could not be measured, such as the cognitive function among the patients at baseline. Fourth, the blood biomarkers were only detected at baseline, and the time of the blood draw is unknown, which would result in no consideration of circadian rhythms and intra-individual variabilities. Fifth, most participants were European in the study, and our conclusions could not be generalizable to other populations worldwide. Finally, the inherent disadvantage of the cohort study could not establish a causal relationship, which indicated reverse causation in the associations between fish oil supplementation and dementia risk might exist in this study. Therefore, the beneficial role of n-3 PUFA supplementation and optimal dosage for older diabetic patients would need to be further explored in future.

5. Conclusions

In summary, fish oil supplementation was related to lower dementia risk among older diabetic patients, whereas both non-oily and oily fish consumption were not associated with dementia risk. The inverse association was not modified by the APOE ε4 genotype. Our findings provide pertinent evidence to support fish oil supplement use for preventing the occurrence of dementia among diabetic patients. Further study is warranted to investigate the effect of fish oil supplementation and the optimal dosage for lowering dementia risk among diabetic patients in future.

Author contribution

Yin Li: Writing - Original Draft, Investigation, Visualization, Formal Analysis, Data Curation. Pan Zhuang: Checking Data, Reviewing. Xiaohui Liu: Checking Data, Reviewing. Shanyun Wu: Reviewing. Yuqi Wu: Reviewing. Lange Zhang: Reviewing. Yu Zhang: Reviewing. Jingjing Jiao: Conceptualization, Supervision, Reviewing, Funding Acquisition.

Funding

The support came from the following foundation: the National Nutrition Science Research Grant of China [grant number: CNS-NNSRG2022-148].

Ethics approval and consent to participate

The ethical application was approved by the research ethics committee (REC reference for UK Biobank 11/NW/0382), and informed consent was offered by participants.

Data availability

This UK Biobank resource was employed to conduct the research under Application Number 47365. More information can be achieved at https://www.ukbiobank.ac.uk.

Consent for publication

Not applicable.

Conflicts of interest

Any conflict of interest did not exist.

Appendix A. Supplementary data

The following is Supplementary data to this article: