Introduction

Over the last decades, the increase in human lifespan and the growing older population worldwide has changed the epidemiology of diseases leading to a substantial rise in age-related health conditions [1, 2]. Mental and cognitive health have been reported to represent an emerging global issue for elderly individuals worldwide [3]. Specifically concerning cognitive-related conditions, Alzheimer’s disease (AD) and other dementias have been estimated to account for about nearly 60 million cases globally projected to rise 3-fold by 2050 [4]. [5, 6] Diet is nowadays recognized to affect human brain and mental health conditions [7, 8]. Several dietary components, such as healthy fats, certain amino acids and oligopeptides, antioxidant vitamins and phytochemicals [i.e., (poly)phenols] are recognized to play a role in preserving neuron stability and functionality as well as counteracting neuroinflammation [9,10,11]. Dietary patterns characterized by fish consumption as one of the main sources of protein [i.e., the Mediterranean diet, the Nordic diet, and the Dietary Approach to Stop Hypertension (DASH)] have been consistently associated with lower risk of neurodegenerative conditions [12,13,14]. Fish has long been studied for its role on human health [15]. Its content in omega-3 polyunsaturated fatty acids (PUFAs) is considered the culprit for the potentially beneficial effects of seafood on mental health [16], while only relatively recently additional attention has been paid to bioactive oligopeptides (bioactive molecules composed of only few amino acids) and their ability to exert direct effects in the brain, demonstrating anti-inflammatory and antioxidant activities [17, 18]. Although the rationale behind the potential beneficial effects of fish intake in neurodegenerative diseases is quite convincing, it is still unclear whether fish consumption per se might play a role in the prevention of cognitive decline and dementia. Two recent meta-analyses explored the role of fish and cognitive outcomes reporting a dose-response association with lower risk of dementia and Alzheimer’s disease [19, 20]: however, the results are outdated, a broader exploration of cognitive outcomes could be further implemented, risk estimates were only provided by weekly intakes, and some missing entries could be integrated. Hence, the aim of the present study was to update current evidence of the association between fish consumption and cognitive decline, impairment, and dementia risk reported in observational studies and provide a summary meta-analysis of the results.

Methods

The design and reporting of this study followed the Meta-analyses Of Observational Studies in Epidemiology (MOOSE) guidelines (Supplementary Table 1) [21]. The systematic review protocol was registered in the PROSPERO International Prospective Register of Systematic Reviews database (ID: CRD42024501232, at https://www.crd.york.ac.uk/prospero/).

Search strategy and study selection

To identify potentially eligible studies, a systematic literature search of PubMed and Scopus databases was performed from their inception up to March 2024. The search strategy was based on the combination of the relevant keywords imputed as text words and MeSH terms, related to fish, seafood and shellfish and cognitive outcomes (Supplementary Table 2). Eligibility criteria for the systematic review and meta-analysis were specified using the PICOS approach (Supplementary Table 3). Studies were eligible if they met the following inclusion criteria: (1) conducted on older adults (i.e., mean age > 50 years old) or, more in general, investigating cognitive outcomes occurring at older age; (2) had observational design (cohort studies, cross-sectional studies, case-control studies); (2) reported exposures to habitual fish, seafood, or shellfish consumption assessed through either 24-h recalls, food frequency questionnaires (FFQ), or dietary diaries; (4) investigated cognitive impairment, cognitive decline, and/or any type of dementia (including Alzheimer’s disease) as outcome; and (5) provided probability measures [odds ratios (ORs), relative risks (RRs), or hazard ratios (HRs)] for the cognitive outcomes investigated. Although the systematic search was not language restricted, only English language studies were eligible. Reference lists of all eligible studies were also examined for any additional studies not previously identified. If more than one study reported results on the same cohort, only the study including the larger cohort size, the longest follow-up, or the most comprehensive data was included in the meta-analysis. The systematic literature search and study selection were performed by two independent authors (J.G. and G.G.) and any incongruity was resolved through a discussion and reaching consensus.

Data extraction and quality assessment

Data from all eligible studies were extracted using a standardized electronic form. The following information was collected: first author name, publication year, study design and location, population age and gender, sample size, details on the assessment method of dietary habits, details on the exposure, details on the assessment method of the outcome of interest, outcome of interest, main findings of the study, measures of association including 95% confidence intervals. The quality of each eligible study was evaluated using the Newcastle-Ottawa Quality Assessment Scale, consisting of 3 domains of quality (selection, comparability, and outcome) and assessing specific study characteristics depending on the type of study design [22]: in general, studies scoring over 5 and 7 points for cross-sectional and prospective studies, respectively, were identified as being of good/high quality. Two investigators extracted the data and assessed the methodological quality independently and any incongruity was resolved through a discussion and reaching consensus.

Statistical analysis

Various risk measures, such as odds ratios (ORs) and hazard ratios (HRs) under the rare disease assumption were treated approximately equivalent to risk ratios and further all were consistently denoted by RRs. The logarithms of RRs from fully adjusted models were pooled in meta-analysis to compare the risk of cognitive events between extreme categories of fish consumption and to reveal dose-dependent relationships. Cognitive impairment and cognitive decline were deemed as a single outcome because, although not clearly stated in all studies providing such outcomes, they both most likely referred to age-related conditions or early-stage disease. All-type dementias and Alzheimer’s disease were investigated as individual separate endpoints. No further data on other specific types of dementia was available in the included studies. RRs for independent studies reported in the same article (i.e., for NHS and HPFS cohorts), were analyzed as separate estimates. When risk estimates were provided for males, females and both sexes together, the latter were used in the main analyses;. when pooled data by sex was not provided in the original study, risk estimates were first pooled using a fixed effect meta-analysis to obtain the joint RR. Der Simonian and Laird random-effects model was applied in which weights of the studies were calculated as the inverse of the sum of both within- and between-study variance [23]. The differences in the research results included in the meta-analysis reflected by the degree of heterogeneity were assessed by the Cochran’s Q-test and the I² index. For Q-test the level of significance was set at p < 0.10 and the value of I2 statistic exceeding 50% was regarded as considerable heterogeneity between studies. A non-linear dose–response meta-analysis was performed only for studies which reported RRs for at least 3 different levels of well-defined fish intake. If the range of fish consumption was not given, the right-unbounded interval was assumed to be the same width as of the adjacent category, while the left-unbounded interval we set to zero. For each category of exposure, the medians, means or midpoints of ranges of daily consumption were extracted directly from the original studies and assigned to the corresponding RRs. When specific quantity of fish intake was not available, daily fish consumption was calculated by multiplying the frequency of consumption (number of serving per day) by the average portion size estimated as 105 g [24, 25]. A dose–response meta-analysis was modeled by restricted cubic splines with the knots at fixed percentiles of the fish intake distribution (10%, 50%, and 90%) [26]. If the distribution of cases and number of participants or person-years was accessible for all categories of fish consumption, we applied the generalized least squares method to estimate trend from summarized dose-response data accounting for the correlation between extracted RRs [27, 28]. Otherwise, a standard technique based on weighted least squares analysis was adapted [27]. The between-study variance-covariance matrices were assessed via multivariate extension of the method of moments to combine all the regression coefficients across studies. A non-linearity was tested by verifying whether the coefficient of the second spline differ from zero. To compare the effects of fish consumption on cognitive function in specific subgroups in which the risk of disease could potentially differ, analyses in strata of APOE genotype (carrying APOE ε4 allele vs. possessing ε2 or ε3 alleles) were performed. Sensitivity and subgroup analysis were also conducted to explore potential sources of heterogeneity. A one-by-one exclusion method was adopted by recalculating combined effect sized after removing one at the time each study. Subgroup analyses were conducted according to year of publication, study quality, age of participants at baseline, study design, length of follow-up, sample size, and type of dementia diagnosis. Small-study effects being the indicator of possible publication bias was examined quantitatively via Egger’s regression test as well as using graphical technique based on visual assessment of asymmetry patterns of funnel plots further adjusting for the number and outcomes of missing studies using trim-and-fill method. R version 4.3.0 (Development Core Team) was used for the statistical analysis. All tests were two-tailed and statistical significance was defined as P < 0.05.

Results

A total of 1169 studies were deemed of potential interest for this systematic search. After removal of 130 duplicates and exclusion of 813 studies through title and abstract evaluation, the full-text from 226 studies were examined. An initial screening was applied based on the following reasons: lack of exposure (n = 85), no outcome of interest (n = 90), different study design (n = 2), only reported on biomarkers of consumption (n = 3), and conducted on younger population (n = 8). The resulting 38 studies were further examined for overlaps. After the final exclusion of 3 studies conducted on the same cohorts, a total of 35 studies [29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64] were included in the present meta-analysis (Supplementary Fig. 1).

Study characteristics exploring fish consumption and cognitive outcomes

The main background characteristics of the studies included, and an overview of the main findings are reported in Table 1. A total of 25 had a prospective design, 8 were cross-sectional and 2 were case-control studies. Ten studies were conducted in Western countries, with 8 specifically involving Northern American cohorts and 13 conducted in European countries, while 13 including Eastern Asian countries. Most studies involved both sexes with just a few exceptions. Among the outcomes investigated through a variety of diagnostic and screening tools, 18 studies explored the relation between fish intake and cognitive decline 15 accounted for diagnosis of dementia, and 11 specifically investigated the risk of Alzheimer’s disease. In general (depending on the differential inclusion in specific analysis), the whole sample included a total of about 849,263 individuals, 8537 comprehensive cases of cognitive impairment/decline, 12,148 cases of dementia, and 5320 cases of Alzheimer’s disease. The quality of the studies scored over 5 for cross-sectional and over 7 for case-control and prospective studies, suggesting an overall good quality of the reports included in this meta-analysis (Supplementary Tables 4–6).

Table 1 General characteristics of included studies
Full size table

Comparison of the risk of cognitive disorders between extreme categories of fish intake

The analysis of the association between fish consumption and dementia, Alzheimer’s disease, and cognitive impairment/decline was based on 15, 10 and 18 studies, respectively (Fig. 1 and Supplementary Table 7). Comparing with the lowest category of fish consumption, the highest consumption was related to 18%, 15% and 18% lower risk of each aforementioned outcome, respectively (RR = 0.82, 95% CI: 0.73–0.93 for dementia, RR = 0.80, 95% CI: 0.67–0.96 for Alzheimer’s disease, and RR = 0.82, 95% CI: 0.75–0.90 for cognitive impairment/decline; Fig. 1). The evidence of substantial heterogeneity was detected for cognitive impairment/decline (I2 = 61%, P < 0.001; Fig. 1). Exclusion of one study at the time did not considerably change any of the results (Supplementary Fig. 2); however, the analysis for cognitive impairment/decline risk resulted in a decrement of I2 statistic to 42% while still maintaining the similar estimate of size effect (RR = 0.76, 95% CI: 0.66–0.88) after exclusion of one study [52]. The inspection of funnel plots and the results of the Egger’s test revealed some evidence of asymmetry in all outcomes except for cognitive impairment/decline (Supplementary Fig. 3): trim-and-fill analysis adjusting for potential publication bias by complementing 5 and 4 missing studies in the case of dementia and Alzheimer’s disease, respectively, confirmed the previous findings (Supplementary Table 7, Supplementary Fig. 4). Subgroup analysis for each outcome showed substantially stable results, with some minor loss of significance in certain subgroups, such as results in European countries in the case of dementia and Alzheimer’s disease, analysis in studies with longer follow-up, and participants below 70 years (Table 2). Additional sub-group analyses focused on prospective studies only showed consistent results with only minor changes (i.e., no significant results in Asian countries, studies with longer follow-up, and larger samples for dementia) (Supplementary Table 8). Notably, most studies on cognitive decline relied on self-reported diagnosis through screening tools, while analyses on dementia and Alzheimer’s disease relied on clinical diagnosis, in both cases reporting stronger reduced risks with the most used diagnostic approach used (Supplementary Table 8).

Fig. 1
figure 1

Meta-analysis of the risk of cognitive outcomes for the highest vs. the lowest fish consumption

Full size image
Table 2 Subgroup analysis for risk of cognitive outcomes according to potential sources of heterogeneity the highest vs. the lowest fish consumption
Full size table

Comparison of the risk of cognitive disorders dependent on the amount of fish consumption

The dose-response analysis using restricted cubic splines is graphically presented in Fig. 2 and RRs are reported in Table 3. A significant decreased risk of cognitive impairment/decline across higher levels of fish intake up to 30% for 150 g/d was found (RR = 0.70, 95% CI: 0.52–0.95), although with large confidence intervals and evidence of significant heterogeneity (I² >90%, P < 0.001). No significant findings were found for dementia and Alzheimer’s disease, although a decreased risk of the latter for up to 50 g/d of fish was reported (Fig. 2and Table 3).

Fig. 2
figure 2

Graphical representation of dose-response meta-analysis of the risk of cognitive outcomes for various servings of fish intake

Full size image
Table 3 Dose-response meta-analysis of the risk of cognitive outcomes for various servings of fish intake
Full size table

Fish intake and cognitive outcomes by APOE genotype strata analysis

A limited number of studies presented results by APOE genotype strata (up to 3 studies, dependent on the endpoint) and results were rather contrasting (Supplementary Fig. 5 and Supplementary Table 9). Results from pooled analyses for dementia resulted in null findings although with a trend toward decreased risk in APO ε2 or ε3 allele carriers (RR = 0.77, 95% CI: 0.58–1.03). However, the only study specifically conducted on cognitive impairment/decline showed a decreased risk for higher fish consumers among APOE ε4 allele carriers (RR = 0.18, 95% CI: 0.05–0.63). No significant associations were found between fish intake and Alzheimer’s disease risk by APOE genotype strata.

Discussion

The aim of this study was to explore the relation between fish consumption and the risk of a variety of cognitive outcomes in observational studies. The main results of the meta-analysis showed that higher fish consumption was associated with lower risk of cognitive impairment/decline, dementia, and Alzheimer’s disease, although a clear dose-response relation could only be observed for the former. A certain degree of heterogeneity could only be partially explained by some variables (i.e., age groups), while differences by genetic background may in fact play a role, yet only limitedly investigated, with not enough studies to effectively draft conclusions on this matter.

Aside from a para-physiological decline in cognitive abilities associated with the growing age, pathological cognitive impairment may depend on a variety of changes in the older brain both determined by genetic and environmental stimuli [65]. Alteration of brain structure, neurotransmission, vascular irroration, and deposit of abnormal proteins (i.e., beta amyloid) are the most common pathological processes determining an alteration of cognitive abilities in older individuals [66,67,68]. Fish is rich in omega-3 PUFA, which have been widely demonstrated to exert a variety of actions in the human brain, including modulating the immune response to insulting stimuli and eventually counteracting neuroinflammation by serving as precursors of pro-resolving mediators, affecting nitric oxide synthesis, decreasing reactive oxygen species (ROS), and more in general improving endothelial dysfunction characterizing certain types of dementia [69,70,71]. Omega-3 PUFA also play an important role in maintaining structural function of the brain, preserving the integrity of the blood-brain barrier, counteract brain atrophy, promoting neurogenesis and increased volume of certain brain area deputed to cognitive functions, the hippocampus [72,73,74]. Although much evidence for such mechanisms is often supported by only preclinical models, current findings match the rationale behind the results from most observational studies conducted so far on fish and cognitive outcomes.

Aside from omega-3 PUFA, recent research has focused on other components of fish that may result in effects on the human brain. Oligopeptides found in fish have been shown to potentially exert neuroprotective effects by serving as precursors of biologically active agents that may counteract some processes occurring in the brain promoting cognitive decline [75]. Some of the mechanisms potentially playing a role against neurodegenerative diseases reported to be exerted by bioactive peptides from seafood include modulation of inflammatory pathways and pro-survival and neurotrophic gene expression, improvement of cell viability, inhibition of acetylcholinesterase and endothelial nitric oxide synthase, and reduction of intracellular antioxidant enzymes depletion [76]. Moreover, inhibitory effects on the beta-secretase enzyme involved in the generation of amyloid-beta peptides that aggregate in the brain of Alzheimer’s patients have also been reported from marine-derived peptides [77]. Although most evidence is yet based on preclinical studies, there is much interest in further investigating the efficacy of such compounds in human in vivo trials. Certain limitations should be considered when exploring their actual capacity to exert effects on human health, including the resistance to digestion operated by proteases and peptidases occurring all over the gastrointestinal tract and the capacity to cross the blood-brain barrier [78]. Nonetheless, the aforementioned mechanisms could support the hypothesis that neuroprotective peptides from fish could play a role against cognitive impairment.

Fish is also a rich source of vitamins and minerals that can play, to a certain extent, a role in brain health [79]. A large variety of minerals, such as iron, magnesium, zinc, phosphorus, and selenium, as well as vitamins, such as group B and D vitamins, are generally well represented in seafood. While there is not much evidence of meaningful effects of supplementation on cognitive decline or dementia [80], all the aforementioned micronutrients are known to play important physiological actions in brain cells, including maintenance of a functional neuroglia, synthesis of precursors of neurotransmitters and control of intracellular calcium release, both important for synaptic excitability and neurotransmission [81]. Nutritional deficiencies lead to documented neurological malfunctioning possibly due to failure of defense mechanisms (i.e., against oxidative stress and inflammation) or age-related frailty, including fatigue and decrease in cognitive performance [82]. Although it is unclear whether the vitamin and mineral content in fish may play a substantial role in preventive dementia and Alzheimer’s disease, they are most likely to affect cognitive abilities and long-term exposure or, on the contrary, chronic deficiency may in fact be an important factor for the maintenance of a healthy brain and decrease the risk of neurodegenerative conditions [83].

The findings on the associating between fish and cognitive outcomes may display a certain degree of heterogeneity across studies because of some variables that should be taken into account when exploring such topic. First, the positive effects of intake of omega-3 toward the central nervous system has been demonstrated to be valid in individuals with cognitive decline or dementia, although the impact on the basic pathological lesion (i.e., amyloid deposition) and more advanced stages of dementia is still unclear [84, 85]. Other factors to be considered to interpret heterogeneity of results include the potential discrepancy between omega-3 PUFA dietary consumption, plasma concentrations, and brain membrane composition, which may be eventually influenced by age or genetic factors [86]. In fact, older individuals may have been suggested to exhibit higher omega-3 PUFA plasma concentrations and yet lower content in their brains [87, 88], leading to a higher susceptibility to potential deficiency and, consequently, stronger effect following exposure. This hypothesis is in line with the results of the present study, being the retrieved association reported especially in individuals older than 70 years old. Among genetic factors, APOE variants (a lipid transporter within the brain) and genes encoding enzymes involved in the leukotriene synthesis, has been shown to interact, albeit with contrasting results, with dietary PUFA intake and their related health effects [89]. Such strata analysis has also been performed in the present meta-analysis and the results pointed to a potential weaker association between fish consumption and APOE ε4 allele, a genetic marker associated with disturbed omega-3 PUFA metabolism leading to lower plasma concentrations than in non-carriers [90, 91], in which significant associations with lower risk of dementia and cognitive impairment were found. Other potential sources of heterogeneity may depend on the type of outcome investigated as well as the type of diagnosis (evaluation through screening tools vs. clinical assessment). For both variables, considering the results reported in this study, we may hypothesize that the risk reduction may occur when considering age-related cognitive decline or generic cognitive deficits and disorders, not just yet developed into well-identified clinical conditions, such as certain types of dementia, including Alzheimer’s disease. Eventually, other factors (i.e., genetic and environmental) may concur to the development of specific neurocognitive disfunctions and diet alone may not be sufficient to actually significantly reduce the risk of their insurgence (at least not observed for fish with the models currently used).

Other limitations potentially determining the heterogeneity of the results (as well as affecting the strength of evidence) include technical and methodological features from the original studies included in the meta-analysis. First, most studies used self-reported dietary information to assess fish consumption, which may be subject to recall bias and social desirability bias. Second, although the quality of the included studies was high, the original study design cannot detect any causal inference, but only associations with risk. Moreover, the adjustment for several potential confounding factors do not guarantee the presence of unmeasured variables potentially playing a role in brain health (i.e., overall diet quality). Ultimately, fish consumption is relatively easy to be estimated and would allow to consider not only the role of omega-3 PUFA but also other components potentially important to exert putative effects on human brain: nonetheless, investigating specific markers in blood or brain would be a further necessary step to increase precision of measurements and inference, ultimately potentially reducing the heterogeneity of findings. Finally, concerning the outcomes, the use of several different tools across the studies may limit the univocity and the consistency of the endpoints investigated.

In conclusion, the present study showed that higher fish intake may be associated with better cognitive status in older individuals. Whether fish consumption may actually decrease the risk of dementia and Alzheimer’s disease is still to be confirmed, but current results are promising. The observed trends of risk estimates suggest a lower risk of disease with increasing consumption of fish starting 50 g per day, while findings for higher intakes are more heterogeneous. The existing mechanistic evidence providing a sound rationale in support of such findings and the consistency of results foster the inclusion of fish in a healthy, balanced dietary pattern. While fish consumption may naturally occur in more coastal areas or, more in general, countries with historical and cultural habits characterized by its inclusion in traditional dietary patterns (i.e., the Mediterranean diet), it is important to consider the importance of food availability and affordability globally. Although evidence from the scientific literature support the mechanical role of omega-3 PUFA in improving certain brain structure, further studies are needed to shed the doubts concerning the actual role of fish intake in more pathophysiological complex conditions, to estimate whether the beneficial effects are in fact exerted by its content in omega-3 or rather other compounds (such as, peptides), and to understand the extent of efficacy also in relation to age and genetic factors. Although interesting and giving the chance for speculation, the findings of this study are yet preliminary and need additional proof to further investigate the role of unmeasured factors, including mechanisms (such as, transportation, membrane incorporation, etc.) related to personalized inter-individual response to omega-3 PUFA or other nutritional compounds retrieved in fish. Also, a more precise clinical characterization of cognitive disorders could help to reduce the heterogeneity of findings and identifying potential specific conditions particularly sensible to the beneficial effects of inclusion of fish in a healthy diet.