Exploring causal correlations between oily fish intake and multiple sclerosis: A two-sample Mendelian randomization study

Abstract

Background and Objectives

According to observational studies, dietary habits may influence the occurrence of multiple sclerosis (MS). There are, however, only a few Mendelian randomization (MR) studies on both.

Methods and Study Design

The objective of this two-sample MR study was to examine possible causal associations between the twenty-one dietary practices and MS. For this investigation, we employed MR analysis utilizing generally accessible statistics from genome-wide association studies (GWAS) to examine causal connections between dietary habits and MS susceptibility among persons of European descent. The IEU Open GWAS project (https://gwas.mrcieu.ac.uk/) provided these GWAS data. Pleiotropy and heterogeneity were investigated using the MR-Egger Intercept test and Cochran's Q test, respectively. MR-Egger, weighted median, inverse variance weighted (IVW), simple mode, and weighted mode were used to assess the causal relationship between 21 dietary habit levels and MS.

Results

After removing outliers, we found a significant association between genetically induced oily fish intake and MS risk (IVW, OR: 0.557; 95% CI: 0.351-0.884; p = 0.013). Extensive sensitivity analyses confirmed this result. Other dietary habits had no discernible relationship with MS risk.

Conclusions

This MR analysis provides evidence of an association between dietary patterns and the risk of developing MS. Notably, higher intake of oily fish was associated with a reduced risk of MS among individuals of European ancestry.

Introduction

Multiple sclerosis (MS) is an autoimmune disorder predominantly involved in the spinal cord and brain.¹ The condition generally becomes apparent upon the initial stages of adulthood and ranks as the primary contributor to young adult-stage non-traumatic neurological impairments.² MS has been on the rise in recent years, though the precise reasons for this remain unknown due to an incomplete understanding of its etiology.

Environmental, nutritional, and lifestyle risk factors, in addition to genetic predisposition, have been shown to contribute to the risk of MS. In recent years, there has been a growing global focus on the potential role of dietary factors in MS risk. Numerous studies have provided substantial evidence that diet can influence both the risk and symptoms of MS. For example, an Iranian case-control study indicated that adhering to a healthy eating pattern may reduce MS risk and an unhealthy eating pattern may lead to increased risk.³ Several other studies have also supported these findings.^{4, 5} Additionally, a recent observational study found that following a healthy eating pattern was linked to a lower risk of MS.⁶

Although research suggests that adopting healthy eating habits can help prevent and manage MS symptoms, studies on the effects of individual foods on MS have produced inconsistent and inconclusive results. Several case-control papers have established higher fish consumption's correlation with mitigated MS risk progression.^{7, 8} However, observational studies have found no significant link between fish consumption and the risk of MS.⁹ Similarly, eating meat products was linked to an increased risk of MS in a factor-analysis study conducted in the United States,¹⁰ but this finding was inconsistent across other studies.^{11, 12} Likewise, some multicenter case-control studies have found that coffee consumption may have a protective effect against MS,¹³ while others have found no significant association.¹⁴ Other potential protective and risk factors identified in observational studies include the consumption of vegetables,^4,12,15,16 fruits,^{4,11-13,17-19} nuts,¹⁷ dairy products,^17,20,21 beans,^{15, 18} cheese,^{12, 15} and alcohol.^{22, 23} It is important to note that nutritional epidemiological studies face challenges in controlling for confounding factors that may influence the relationship between diet and MS. Moreover, most previous dietary trials in MS have lacked methodological rigor. Therefore, robust evidence is needed to demonstrate whether dietary behaviors can prevent MS onset or modify its progression.

As an epidemiological methodology, Mendelian randomization (MR) employs the distinctive genotype characteristics to examine the causal connection between exposure and outcomes.^{24, 25} By employing instrumental variables (IVs) as genetic variants that are strongly correlated with exposure and are allocated randomly to populations during meiosis as well as conception, it emulates an environment with randomized controls. Potential confounding variables and reverse causal bias can both be circumvented using MR designs. A number of genome-wide association studies (GWAS) have uncovered evidence of inherited dietary habits.^{26, 27} Consequently, MR can serve as a suitable research design to evaluate the influence of dietary habits on health outcomes or diseases.²⁸

By leveraging extensive GWAS data and MR analysis, we examine the associations between 21 dietary habits and the risk of MS, aiming to identify specific diets that may have a causal effect. This study is among the first to investigate the potential causal relationship between dietary practices and MS risk using MR analysis. Our findings may offer a scientific basis for MS prevention strategies.

Methods

Study design

We examined the causal association of exposure against the results by employing genetic variants to be the IVs in this two-sample MR study. Figure 1 illustrates our MR design flowchart. The Strengthening the Reporting of Observational Studies in Epidemiology using Mendelian Randomization (STROBE-MR) checklist was completed for this observational study (Supplementary 1).

Figure 1.

The MR study flowchart that reveals the causality between dietary habits and MS risk. MR, Mendelian randomization; MS, multiple sclerosis

The selection of dietary exposures in this study was based on two fundamental criteria to ensure both scientific rigor and biological relevance. First, all dietary traits included had to have available GWAS summary statistics from the UK Biobank (UKB), a large, prospective population-based cohort comprising approximately 500,000 participants with extensive genetic and phenotypic data. The UKB dataset is characterized by standardized data collection procedures and stringent quality control measures, which together guarantee the accuracy and reliability of genetic associations with phenotypes. Utilizing high-quality GWAS data allows for the construction of strong genetic instruments, minimizing weak instrument bias and enhancing the statistical power of MR analyses. Second, the selected dietary exposures were required to possess clear biological plausibility, supported by prior epidemiological, clinical, and experimental evidence implicating these factors in the pathogenesis of MS. Specifically, dietary components such as polyunsaturated fatty acids, red meat consumption, and alcohol intake have been shown to influence immune regulation, inflammatory responses, and metabolic pathways relevant to MS development. By integrating evidence from these diverse sources, this study ensures that the chosen exposures are both mechanistically relevant and amenable to robust genetic analysis, thereby strengthening the causal inference regarding diet and MS risk.

Exposure GWAS

This study included a total of 21 dietary exposure variables encompassing a broad spectrum of commonly consumed foods and beverages, which were systematically categorized into five main groups: meat intake (processed meat [N=13,150], pork [N=8,630], lamb/mutton [N=20,900], poultry [N=5,470], and beef [N=1,910]); fish intake (non-oily fish [N=4,458] and oily fish [N=95,507]); beverages and related habits (tea [N=37,189], coffee [N=15,854], weekly alcohol intake [N=26,805], alcohol intake frequency [N=169,927], water intake [N=49,416], and hot drink temperature [N=39,738]); staple foods (cereal intake [N=14,330] and bread intake [N=80,777]); and fruits, vegetables, and other foods (dried fruit [N=14,501], fresh fruit [N=62,710], salad/raw vegetable intake [N=7,252], cooked vegetable intake [N=59,215], added salt intake [N=79,294], and cheese intake [N=45,541]). The GWAS summary statistics for these dietary traits were obtained from the UK Biobank (https://www.ukbiobank.ac.uk), which provides detailed information on sample sizes, phenotype definitions, and stringent quality control procedures, thereby ensuring the reliability and consistency of the data. Further specifics, including detailed phenotype descriptions and statistical metrics, are provided in Supplementary Tables 2 and 3. This comprehensive and systematic selection of dietary exposures enables robust Mendelian randomization analyses to explore the potential causal relationships between diet and multiple sclerosis risk.

Outcome GWAS

We developed an instrumental genetic proxy for MS using information taken from the more extensively publicized GWAS from the International Multiple Sclerosis Genetics Consortium (IMSGC). The IMSGC included 68,374 controls and 47,429 MS cases, all European descent.²⁹

Selection of instrumental variables in dietary habits

IVs consisting of genetic variants were employed in the MR analysis to examine the causal connection of the exposure against the outcome. Single nucleotide polymorphisms (SNPs) were most prevalent among these IVs. To determine which IVs met the criteria, we devised a sequence of procedures for quality control. First, SNPs significantly associated with each of the 21 dietary habits were chosen as IVs (p <5×10^-8). Next, to assure IV independence, we establish linkage disequilibrium (LD) limits of aggregation distance >10,000 kb and r² <0.001. Then, we eliminated palindromic SNPs that exhibit intermediate allele frequencies and SNPs that were absent from the outcome dataset from each MR analysis. Fourth, each SNP was carefully reviewed in the PhenoScanner database (http://www.phenoscanner.medschl.cam.ac.uk/) to ensure unconfused instrumental variables were only associated with outcomes through associated exposures.

We carefully checked each SNP to see if it was linked to any known confounders that could affect MS risk, using a strict significance cutoff (p <5×10⁻⁸). The analysis adjusted for multiple confounders including age at menarche, body mass index (BMI), obesity status, smoking habits, and socioeconomic status, as these factors may influence MS risk through immune regulation and inflammatory mechanisms. We used publicly available GWAS data to identify and exclude any SNPs significantly associated with these confounders, helping to avoid bias from horizontal pleiotropy.

This filtering step is key to meeting the core assumptions of MR, ensuring that our genetic instruments are not related to confounders and only impact MS risk through the specific dietary exposures under study. By removing SNPs that might violate these assumptions, we reduce potential bias and strengthen the confidence in our causal conclusions. Overall, this quality control process ensures that the links we find between diet and MS risk are more likely to reflect true cause-and-effect relationships, rather than being driven by confounding factors or pleiotropy. Supplementary 4 displays detailed summary statistics for these included SNPs. In addition, F-values were calculated to evaluate weak tool bias using the formula (N−2) × R²/(1−R²), where N stands for sample size and R2 refers to the variance in dietary habits explained by the genetic instrument.³⁰

MR analysis

R 4.2.2 (R Foundation for Statistical Computing, Vienna, Austria) was utilized to conduct MR analyses, specifically the packages TwoSampleMR and Mendelian Randomization Pleiotropy RESidual Sum and Outlier (MR-PRESSO). Inverse variance weighted (IVW) regression was employed as the principal analysis approach to ascertain the possible correlation of dietary habits with MS. This was due to the fact that IVW regression computed the exposure causation on the outcomes by means of Wald estimate of the SNP ratio related to exposure. When horizontal pleiotropic effects were absent from the IVs, the IVW method yielded the most dependable outcomes.

Sensitivity analysis

The primary analyses’ robustness was evaluated through the implementation of six analyses for sensitivity. The aforementioned items were as follows: (i) weighted median regression that yields unbiased slope estimates when a minimum of 50% SNPs are valid IVs;³¹ (ii) weighted mode regression that yields unbiased slope estimates when a multitude of SNPs are valid IVs;³² (iii) Egger regression that examines horizontal pleiotropy (global directional);³³ (iv) simple mode for supplementary examination; (v) MR-PRESSO that seeks IVs depicting horizontal pleiotropy and undertakes slope recalibration following the removal of those searched variables;³⁴ and (vi) leave-one-out IVW regression, which examines the results’ robustness by excluding individual instrumental variables. p < 0.05 denotes a statistically significant outcome. The Cochrane Q’ test, a standard method for MR analysis, was used to determine heterogeneity. Pleiotropy was investigated by utilizing the MR-Egger intercept.

Results

The sample sizes for dietary practices ranged from 335,394 to 462,346 participants. Summary-level data for MS were obtained from IMSGC, comprising 47,429 MS cases and 68,374 controls, with minimal overlap between the exposure and outcome populations. SNPs associated with dietary practices were identified using the ‘harmonise_data’ and ‘extract_outcome_data’ functions to ensure validity and independence. Supplementary 2 provides details on the SNPs linked to the 21 dietary practices. Between 4 and 65 SNPs were deemed suitable as instrumental variables. All selected SNPs had F-statistics greater than 10, indicating a low risk of weak instrument bias.

Our analysis identified a protective association between higher oily fish intake and a reduced risk of MS. Initial MR analyses indicated no significant association between oily fish intake and MS risk(IVW, OR = 0.603, 95%CI: 0.309-1.177, p = 0.138; Supplementary 5), although substantial heterogeneity was observed (p = 4.556×10^-7; Supplementary 6). Nevertheless, after correcting for outlier SNPs using the MR-PRESSO method, the association reached statistical significance (IVW, OR = 0.557, 95%CI: 0.351-0.884, p = 0.013). Comparable results were obtained via the weighted median method following the exclusion of outliers IVW, OR = 0.513, 95%CI: 0.266-0.988, p = 0.046), further underscoring the robustness of the association. Tests for directional pleiotropy and residual heterogeneity, including Cochran's Q statistic and the MR-Egger intercept, provided no evidence of bias after removing outlier SNPs (Supplementary 7). The forest plot (Figure 2A) displays the association estimates for each SNP linked to oily fish consumption and MS risk. The leave-one-out analysis (Figure 2B) indicates that no single SNP exerted disproportionate influence on the overall causal estimate. The scatter plot (Figure 2C) visualizes SNP-specific estimates across multiple MR approaches, while the funnel plot (Figure 2D) did not reveal evidence of bias.

Figure 2.

The connection between the SNP associated with oily fish consumption and MS risk. (A) A forest plot illustrating a two-sample MR analysis; the increase in oily fish consumption corresponds to a higher log-to-value OR of MS, as denoted by the black dots. The causal estimates for every SNP combination, as computed through diverse MR algorithms, are denoted by the red dots. (B) Results of a “leave-one-out” sensitivity analysis; the log-to-value OR of MS rises with oily fish ingestion, as indicated by the black dots. The causal estimates for every SNP combination, as computed through diverse MR algorithms, are denoted by the red dots. (C) A scatter diagram illustrating the results of a two-sample MR analysis; the line's slope denotes the causal estimates of the MR approach, and each black dot corresponds to a single SNP. (D) A funnel diagram representing MR analysis involving two samples; β denotes the regression coefficient and SE the standard error. MR, Mendelian randomization; MS, multiple sclerosis; SNP, single nucleotide polymorphisms; OR, odds ratio; SE, standard error

Except for fish oil, none of the evaluated dietary exposures exhibited statistically significant associations with MS, regardless of outlier exclusion. Specifically, processed meat intake (IVW, OR = 0.801, 95%CI: 0.347-1.849, p = 0.603), pork (IVW, OR = 3.415, 95%CI: 0.617-18.890, p = 0.159), lamb/mutton (IVW, OR = 0.809, 95%CI: 0.325-2.009, p = 0.647), poultry (IVW, OR = 1.353, 95% CI: 0.158-11.595, p = 0.783), beef (IVW, OR = 0.264, 95%CI: 0.060-1.163, p = 0.078), and non-oily fish (IVW, OR = 0.798, 95% CI: 0.215-2.956, p = 0.735) showed no significant associations. Similar null findings were observed for tea consumption (IVW, OR = 0.759, 95%CI: 0.430-1.339, p = 0.340; after outlier removal: IVW, OR = 0.665, 95%CI: 0.406-1.089, p = 0.105), coffee (IVW, OR = 0.503, 95%CI: 0.247-1.024, p = 0.058), and alcohol intake measured both as drinks per week (IVW, OR = 1.046, 95%CI: 0.517-2.115, p = 0.901) and frequency (IVW, OR = 1.314, 95%CI: 0.889-1.943, p = 0.171; post-outlier removal: IVW, OR = 1.277, 95%CI: 0.977-1.671, p = 0.074). Other dietary factors, including water intake (IVW, OR = 0.916, 95%CI: 0.389-2.159, p = 0.842), hot drink temperature (IVW, OR = 1.578, 95%CI: 0.760-3.277, p = 0.221), cereal (IVW, OR = 0.955, 95%CI: 0.449-2.029, p = 0.904), bread (IVW, OR = 0.713, 95%CI: 0.166-3.054, p = 0.648; post-outlier removal: IVW, OR = 0.540, 95%CI: 0.249-1.173, p = 0.120), dried fruit (IVW, OR = 0.740, 95%CI: 0.362-1.513, p = 0.410), fresh fruit (IVW, OR = 0.789, 95%CI: 0.273-2.279, p = 0.662; post-outlier removal: IVW, OR = 0.760, 95%CI: 0.311-1.860, p = 0.549), raw vegetables or salad (IVW, OR = 0.948, 95%CI: 0.171-5.271, p = 0.951), and cooked vegetables (IVW, OR = 7.917, 95%CI: 0.312-201.014, p = 0.210; post-outlier removal: IVW, OR = 1.530, 95%CI: 0.431-5.434, p = 0.511), as well as salt (IVW, OR = 1.057, 95%CI: 0.735-1.520, p = 1.130) and cheese intake (IVW, OR = 1.038, 95%CI: 0.627-1.717, p = 0.886), likewise demonstrated no robust association with MS risk. Despite the presence of variability in the water intake exposure (Cochran's Q test p <0.05), the MR-Egger intercept outcome did not indicate any directional pleiotropy. Our MR study's findings from are summarized in Supplementary 5-7 and Figure 3-4.

Figure 3.

Forest plot using the IVW (inverse variance weighted), MR-Egger and MR-PRESSO to visualize the causal effects of dietary habits (processed meat intake, pork intake, lamb/mutton intake, poultry intake, beef intake, non-oily fish intake, oily fish intake, tea intake, coffee intake, alcoholic drinks per week, alcohol intake frequency) on MS. IVW, inverse variance weighted; MR-PRESSO, Mendelian Randomization Pleiotropy RESidual Sum and Outlier; OR, odds ratio; CI, confidence interval

Figure 4.

Forest plot using the IVW (inverse variance weighted), MR-Egger and MR-PRESSO to visualize the causal effects of dietary habits (water intake, hot drink temperature, cereal intake, bread intake, dried fruit intake, fresh fruit intake, salad/raw vegetable intake, cooked vegetable intake, salt, cheese intake) on MS. IVW, inverse variance weighted; MR-PRESSO, Mendelian Randomization Pleiotropy RESidual Sum and Outlier; OR:odds ratio; CI, confidence interval

Discussion

The most significant finding of this study was the association between oily fish consumption and MS risk. To address potential outliers, we repeated the MR analysis using revised IVW estimates. After excluding outliers, a causal association was identified: higher oily fish intake was linked to a reduced risk of MS. These results may support clinicians in promoting dietary changes—such as increased consumption of oily fish—as part of health education for MS patients and individuals at high risk. Overall, this study contributes important insights into both risk and protective dietary factors for MS and represents one of the most comprehensive MR analyses to date examining the potential causal role of diet in MS development.

A growing body of evidence supports an association between fish consumption and reduced risk of developing MS. For example, a study by Kampman et al. among adolescents found that consuming fish at least three times per week was associated with a lower likelihood of developing MS (OR = 0.55, 95%CI: 0.33-0.93, p = 0.024).⁸ According to a systematic review, increased fish intake was linked to reduced MS progression, with protective effects observed at a minimum intake of 0.5 servings per week.³⁵ Additionally, several studies have examined the effects of different fish types on MS prevalence. An Australian study reported that consuming two servings of canned (oily) fish per week was associated with a ~40% reduction in the risk of focal cortical dysplasia (FCD), a condition frequently seen prior to MS onset.³⁶ Other research has also found an inverse relationship between high-fat fish consumption and MS incidence.⁷ However, few studies have established a definitive association between fish intake and MS risk.⁹ Inconsistencies across studies may be due to differences in design and methodology, such as variations in food frequency questionnaires, evolving diagnostic criteria for MS, and residual confounding.

Several foundational studies support our findings. Omega-3 polyunsaturated fatty acids (n-3 PUFAs) have been identified as important modifiable factors associated with MS risk.³⁷ The primary dietary source of n-3 PUFAs is oily fish. Animal studies have demonstrated that n-3 PUFAs exert neuroprotective effects during aging,³⁸ and suppress MS-related inflammation through multiple mechanisms.^{39, 40} These fatty acids have been shown to possess anti-inflammatory properties by modulating transcription factor activity, intracellular signaling pathways, and gene expression.⁴¹ A recent review also reported that n-3 PUFA and fish oil supplementation can reduce relapse rates, lower inflammatory markers, and improve quality of life in individuals with MS.⁴² However, some researchers attribute the potential protective effect of oily fish to its vitamin D content, which may reduce MS risk via systemic circulation.^{43, 44} In summary, oily fish provides both vitamin D and very long-chain omega-3 polyunsaturated fatty acids (VLCn-3 PUFAs),^{45, 46} both of which may contribute to reduced MS susceptibility.

This MR study offers several strengths. First, we assessed the causal relationship between dietary behaviors and MS using a two-sample MR approach, which minimizes the influence of potential confounding factors. Second, we leveraged large-scale GWAS data, including summary statistics from 47,429 MS cases and 68,374 controls, along with extensive datasets on dietary habits. Third, population stratification bias was reduced by focusing on individuals of European ancestry in the GWAS datasets. Fourth, we identified valid instrumental variables using multiple robust statistical methods based on independent genetic variants. Finally, the associations observed were consistent, with no evidence of significant heterogeneity or horizontal pleiotropy. Nonetheless, causality should be interpreted with caution, and further studies are necessary to replicate and generalize these findings.

This study also has several limitations. First, our MR findings are based on populations of European ancestry, which limits the generalizability of the results to other ethnic groups. Caution should be exercised when interpreting these findings in diverse populations. Second, the biological functions of many SNPs used as instrumental variables remain unclear. Third, due to limitations in the available data, we were unable to investigate the effects of comprehensive dietary patterns, such as the Western or Mediterranean diets. Finally, dietary intake data were collected retrospectively using a brief food frequency questionnaire, making it difficult to eliminate potential recall bias and seasonal variation.

Graphical abstract

Conclusions

In summary, we conducted a systematic investigation into the potential causal relationship between dietary habits and the risk of MS. Our MR analysis suggests that higher consumption of oily fish is causally associated with a reduced risk of MS. By highlighting the potential role of dietary factors in MS onset, our findings may inform the development of more targeted and evidence-based prevention strategies.

Supplementary Materials

All supplementary tables and figures are available upon request to the editorial office. GWAS for MS can be found in IMSGC.

Acknowledgements

The authors declare no conflict of interest.

This study did not receive any funding.

Conflict of Interest and Funding Disclosures

No competing interests are reported.