Relationship between Polyunsaturated Fatty Acids and Inflammation: evidence from cohort and Mendelian randomization analyses

Daisy C. P. Crick, Sarah Halligan, George Davey Smith, Golam M. Khandaker, Hannah J. Jones
doi: https://doi.org/10.1101/2023.11.27.23299099

Abstract

Dietary polyunsaturated fatty acids (PUFAs) are thought to influence the risk of various chronic diseases by modulating systemic inflammation. Omega-3 and omega-6 FAs are thought to have anti- and pro-inflammatory roles, respectively, but it is unclear whether these associations are causal. We tested associations of PUFAs with three blood-based biomarkers of systemic inflammation, namely C-reactive protein (CRP), glycoprotein acetyls (GlycA) and interleukin 6 (IL-6), in a population cohort (n=2748) and using Mendelian randomization analysis (a genetic causal inference method). We provide consistent evidence that omega-6 PUFAs increase GlycA levels, but omega-3 FAs do not lower levels of inflammatory markers. Additionally, we found that a higher omega-6:omega-3 ratio increases levels of all three inflammatory markers; CRP (mean difference=0.17; 95% CI=0.13, 0.20), GlycA (mean difference=0.16; 95% CI=0.13, 0.20) and IL-6 (mean differene=0.19; 95% CI=0.15, 0.22) in the cohort analysis. Our findings suggest that future public health messaging should encourage reducing the consumption of omega 6 FAs and maintaining a healthy balance between omega 3 and omega 6 FAs, rather than focusing on omega-3 FA supplementation. This is because dietary omega-3 supplementation alone is unlikely to help reduce systemic inflammation or inflammation-related disease.

1. Introduction

Non-communicable diseases (NCDs) are the main cause of global mortality, representing 71% of deaths worldwide2. As such, their prevention remains a key challenge. Systemic low-grade inflammation, reflected by elevated concentrations of markers of inflammation such as cytokines (e.g. interleukin-6 (IL-6)) and acute phase proteins (e.g. C-reactive protein (CRP)), has been implicated in the pathophysiology of many NCDs from cardiovascular disease to depression3–12. Therefore, inflammation appears a promising target for their treatment and prevention.

Polyunsaturated fatty acids (PUFAs), are proposed to influence levels of systemic inflammation13. There are two main families of PUFAs which are essential for many metabolic processes: Omega-3 (n-3)14, 15; and Omega-6 (n-6)15, along with eicosatetraenoic acid and docosahexaenoic (DHA) acid (both n-3 PUFAs), and arachidonic acid (ARA; an n-6 PUFA) linked most clearly to various essential physiological processes. These processes include maintaining cell membrane structural integrity, hormone synthesis, regulation of gene expression, and supporting brain and nervous system functioning16. Some PUFAs such as ARA can come from both exogenous sources (such as meat and egg products) and endogenous sources. This is where the PUFAs can be synthesized from their metabolomic precursors: α-linolic acid (ALA; an n-3 PUFA) and linolic acid (LA: an n-6 PUFA), through desaturation and elongation reactions17. In comparison, humans cannot synthesise LA or ALA and therefore they must be acquired through dietary consumption18 such as vegetable oils (for LA) and seed oils (ALA).

There is some evidence that N-3 PUFAs have anti-inflammatory effects and protect against the severity/occurrence of inflammation-related conditions19, 20. For example, there are suggestions that fatty fish consumption can reduce the probability of death from a heart attack 21, 22. This is attributed to n-3 PUFAs ability to alter the production of prostaglandins, which subsequently reduce inflammation (measured by the reduced concentration of circulating inflammatory biomarkers in the blood)23–26. However, clinical trials investigating the effects of n-3 supplementation have yielded contradictory results 27 28 regarding its effect on systemic inflammation. Dietary supplementation studies have reported that the consumption of n-6 PUFAs, which are typically thought to be pro-inflammatory, did not affect the concentration of inflammatory markers such as IL-6 or CRP 23, 29–31.

A meta-analysis of 14 clinical trials that included total 1,35,291 participants found that omega-3 supplementation reduced the risk of major adverse cardiovascular event, cardiovascular death and myocardial infarction 32. In contrast, an RCT study investigating dietary intake and mortality in 3114 men with angina, found that risk of cardiac death was higher among individuals advised to eat oily fish or to take fish oil compared to those not advised to do so. Additionally, RCTs investigating supplementation in patients to treat other inflammation-related NCDs found similarly inconsistent results (e.g. 33 34). However, these studies did not directly test the effect of fatty acids on inflammation levels and the inconsistent findings of the RCTs cast doubt on the posited immune modulatory effect sof n-3 and n-6. Further, the contradictory results also raise questions of whether PUFAs causally influence concentrations of biomarkers of inflammation or whether the observed associations are the result of residual confounding.

Both n-3 and n-6 PUFAs are metabolized by the same enzymes and compete for desaturation and elongation. This means it is plausible that n-6 PUFAs act as competitive inhibitors of n-3 PUFAs and thus reduce the amount of end-product n-3 PUFAs that can be synthesised35. As a result, a lower concentration of n-3 PUFAs could reduce their anti-inflammatory effects36–38. In contrast, a lower plasma n-6:n-3 ratio may result in higher levels of end-product n-3 PUFAs which would subsequently lower concentrations of inflammatory markers leading to beneficial effects on health outcomes39. As such, further research to understand the importance of the n-6:n-3 ratio on levels of inflammation and its impact on the occurrence of NCDs is of public health concern.

We examined whether circulating levels of PUFAs are associated with systemic biomarkers of inflammation using (1) data from a population-based prospective birth cohort and (2) two-sample Mendelian randomization (MR), a causal inference method that uses genetic variants as proxies to examine whether the association between a risk factor and a disease outcome is likely to be causal or result of residual confounding or reverse causaltion40.

We investigated the effects on inflammatory markers of specific n-3 and n-6 PUFAs which play a role in key metabolic processes. These are the long-chain n-3 PUFA DHA and the short-chain n-6 PUFA LA which appear at opposite ends of the biosynthesis pathways (see figure 1; adapted from Videla et al.). In addition, for a greater interrogation of the PUFA-inflammation relationship, we used measures of total n-3 PUFAs, total n-6 PUFAs and the ratio of total n-6 PUFAs:total n-3 PUFAs (referred hereafter as the total n-6:n-3 ratio) as exposures. As outcomes we used three biomarkers for systemic inflammation, namely CRP, IL-6, and Glycoprotein Acetyls or GlycA (a novel composite biomarker of inflammatory acute phase proteins41, 42).

Figure 1:
Figure 1:

Schematic representation of polyunsaturated fatty acid biosynthesis in mammals adapted from Videla 20221

2. Results

2.1 Findings from Population-based Cohort Analyses

Table 1 presents the median/range for PUFA levels (DHA, LA, total n-3 PUFAs and total n-6 PUFAs) and inflammatory biomarker levels (IL-6, CRP and GlycA) from participants of the Avon Longitudinal Study of Parents and Children (ALSPAC) birth cohort at age 24y. There was no difference between the complete-case and imputed results (eTable 15) and distributions of observed and imputed characteristics are presented in supplementary eTable 16. The results of the below associations are presented in eTable17 and eTable 18.

View this table:
Table 1: Median and interquartile range of exposure and outcome data (n = 2748)

2.1.1 Associations of DHA and LA with inflammatory markers

Both DHA and LA were associated with higher CRP and GlycA levels, but lower IL-6 levels at 24y after adjusting for potential confounders relating to sex, substance use, social economic position, and maternal pregnancy health (Figure 2).

Figure 2:
Figure 2:

Association between fatty acids and inflammatory biomarkers using cohort data after adjusting estimates for household social class, maternal highest education qualification, maternal and paternal smoking status during pregnancy, offspring sex and age and status at 24 year clinic

2.1.2 Associations of total n-3 and n-6 PUFAs with inflammatory markers

Total n-3 and n-6 PUFAs were associated with higher GlycA levels, but lower IL-6 levels at 24y after adjusting for potential confounders. Total n-6 was associated with higher CRP levels but there was no strong evidence of association between total n-3 PUFAs and CRP (Figure 2).

2.1.3 Associations between total n-6:n-3 ratio and inflammatory markers

The total n-6:n-3 ratio was associated with higher levels of all three inflammatory markers (CRP, GlycA and IL-6) at age 24y after adjusting for potential confounders (Figure 2).

2.1.4 Sex Specific Associations between PUFAs and Inflammatory Markers

In sex-stratified analyses, the n-6:n-3 ratio was associated with higher levels of all three inflammatory markers in both sexes (Figure 3). PUFAs were also associated with higher GlycA in males and females. The effects of all PUFAs on CRP and IL-6 levels attenuated to the null in males. In females, PUFAs were associated with higher CRP but lower IL-6 levels, which mirrored findings from the whole cohort analysis.

Figure 3:
Figure 3:

Association between fatty acids and inflammatory biomarkers stratified by sex using cohort data after adjusting estimates for household social class, maternal highest education qualification, maternal and paternal smoking status during pregnancy, offspring sex and age and status at 24 year clinic

2.2 Results for Mendelian Randomization Analyses

We used two-sample MR to assess evidence for causality using publicly available summary-level data from European population GWAS of the PUFAs43, GlycA44, CRP45 and IL-646. See eMethods for further information.

Tests investigating instrument validity, described in the methods section, indicated that instruments were unlikely to be subject to weak instrument bias (see eResults).

2.2.1. Potential causal effect of DHA on inflammatory markers

We observed no strong evidence of effect of DHA levels on CRP (IVW estimate: 0.03; 95% CI: -0.01, 0.09), GlycA (IVW estimate: 0.01; 95% CI: -0.06, 0.07) or IL-6 (IVW estimate: 0.07; 95% CI: -0.20, 0.07). These results were largely consistent across sensitivity analyses using MR methods that make different assumptions about instrument validity (MR-egger, weighted median and weighted mode) (Figure 4).

Figure 4:
Figure 4:

Univariable causal effect of fatty acids on Inflammation

2.2.2. Potential causal effect of LA on inflammatory markers

We observed no strong evidence of effect of LA levels on CRP (IVW estimate: -0.10; -0.10, 0.09) or IL-6 (IVW estimate: 0.02; -0.05, 0.08). We observed estimates suggesting that higher LA levels cause higher GlycA levels (IVW estimate: 0.25; 0.15, 0.35). Other than in the analysis investigating the effect of LA levels on CRP, results were consistent across MR sensitivity methods (Figure 4).

2.2.3. Potential causal effect of total n-3 PUFA levels on inflammatory markers

Higher total n-3 PUFA levels were associated with higher CRP (IVW estimate: 0.09; 0.03, 0.16) and GlycA (IVW estimate: 0.12; 0.04, 0.21) levels. We observed no effect of total n-3 PUFA levels on IL-6 (IVW estimate: -0.04; -0.09, 0.02). Results attenuated to the null in all MR sensitivity analyses of total n-3 PUFA levels and GlycA (Figure 4).

2.2.4. Potential causal effect of total n-6 PUFA levels on inflammatory markers

We observed no strong evidence of effect of total n-6 PUFA levels on CRP (IVW estimate: 0.04, -0.05, 0.13) or IL-6 (IVW estimate: -0.08; -0.03, 0.04). We did observe estimates that suggest that higher n-6 PUFA levels cause higher GlycA levels (IVW estimate: 0.25; 0.16, 0.34) with estimates consistent in sensitivity analyses with the exception of the weighted mode analysis (Figure 4).

2.3 Evidence of pleiotropy and heterogeneity

MR Egger intercepts and tests for heterogeneity between the SNP effect estimates are presented in eTable 19. There was no strong evidence of pleiotropic effect as detected by the MR-Egger intercept for the association between PUFA levels and any biomarker with the exception of the association between total n-3 levels and GlycA. Findings were fairly consistent with results using the MR-PRESSO test for pleiotropy (eTable20). There was strong evidence of heterogeneity in all analyses apart from in analyses of total n-3 and LA levels and IL-6.

2.4 Findings on the potential impact of methodological biases in MR results

MR-Lap 47was used to assess potential biases in observed effect estimates due to sample overlap between the fatty acid, CRP and GlycA GWAS. Results suggested that the majority of analyses were largely unaffected by winner’s curse and weak instrument biases introduced through sample overlap. The corrected effect differed from the observed effect only for analysis investigating the effect of total n-6 PUFAs on GlycA and CRP levels. In both instances, the observed effect estimate was in the same direction as the corrected estimate and the confidence intervals did not cross the null but were wider. See eResults for further information.

2.5 Results from MR Analysis Focusing on Specific Genes

The FADS gene cluster and ELOVL2 gene encode key desaturase and elongase enzymes respectively and are involved in the n-3 and n-6 fatty acid biosynthesis pathways. Therefore, we conducted a complementary and mechanistically informative MR analysis by using only genetic instruments from within or close to the FADS gene cluster (FADS1, FADS2 and FADS3) and ELOVL2 gene with the aim of reducing the impact of other pleiotropic pathways. MR analyses using SNPs from the FADS gene region showed that DHA increases CRP levels (IVW estimate: =0.04; 95% CI=0.01, 0.07), but LA decreases CRP levels (IVW estimate: =-0.14; -0.20, -0.07). FADS instrumented DHA and LA had no effect on GlycA and (where testable) IL-6. There was no evidence of a causal effect of any SNPs from the ELOVL2 gene on GlycA, CRP, or IL-6. See eResults for further information.

2.6 Results from analyses exploring heterogeneity of MR instruments

Given that evidence of associations between the PUFAs and GlycA were largely consistent across cohort and MR analyses, we used MR-Clust 48to investigate whether there were distinct clusters of SNPs effects driving the heterogeneity observed in the MR analysis with GlycA levels as the outcome. In the analysis using genetic instruments for LA, total n-3 PUFAs and total n-6 PUFAs, all clusters were positively associated with GlycA which is consistent with the main results. However, for DHA, there were two SNP clusters showing a positive association with GlycA and one cluster showing a negative association with GlycA levels. This could explain the observed null results between DHA and GlycA found in the main analysis. See eResults, eTable 26, and eFigures 3-6 for further information.

We obtained functional annotations of the genomic loci associated for all identified clusters from FUMAGWAS (https://fuma.ctglab.nl) and investigated in which tissues the genes were up- or down-regulated (eFigures 7-10). Overall, most differentially expressed genes were expressed in the liver. For the association between DHA and GlycA, tissues with most differentially expressed genes were: liver (cluster 1), breast (cluster 2) and kidneys (cluster 5). For the association between LA and GlycA, tissues with most differentially expressed genes were: liver (clusters 1 and 2) and kidneys (cluster 4). For the association between total omega-3 and GlycA, most differentially expressed genes in both clusters 2 and 6 were in liver. Finally, for the association between total omega-6 and GlycA, most differentially expressed genes were in kidneys (cluster 1) and liver (cluster 4).

2.7 Multivariable MR investigating the direct effects of total omega-3 and total omega-6 on inflammatory markers

We conducted a Multivariable Mendelian Randomization (MVMR) analysis to estimate the individual direct causal effects of total n-3 and n-6 PUFAs on the biomarkers of inflammation independently of each other. We observed a positive direct effect of total n-6 PUFA levels on both GlycA (IVW estimate: 0.33; 0.19, 0.47) and CRP levels (IVW estimate: 0.20; 0.01, 0.39), but not IL-6 levels (IVW estimate: 0.17; -0.17, 0.50), after controlling for the effect of n-3 PUFAs. There was no direct effect of total n-3 PUFA levels on GlycA (IVW estimate: 0.06; -0.05, 0.16), CRP (IVW estimate:0.03; -0.06, 0.12) or IL-6 (IVW estimate: -0.18; -0.49, 0.14) levels after controlling for the effect of n-6 PUFAs. This suggests that the effect of total n-3 PUFAs on CRP levels found in the univariable MR analysis are not independent of the effect of n-6, and may be a result of pleiotropy (Figure 5). See eResults for further information on MVMR analysis.

Figure 5:
Figure 5:

Multivariable Mendelian randomization analysis of direct effect of Omega-3 and Omega-6 on Inflammation

3. Discussion

By combining complementary cohort and genetic analyses, we provide greater insight into the relationships between dietary PUFAs and systemic inflammatory markers. Our analyses of population cohort data show a consistent effect of the n-6:n-3 ratio on higher levels of all three inflammatory markers in the total sample and in males and females separately. However, the picture appears to be more complex when we consider total or individual FA levels. For instance, total n-3 FAs levels which are thought to be anti-inflammatory are not associated with CRP levels. Similarly, n-6 (thought to be pro-inflammatory) as well as n-3 PUFAs are associated with an increase in GlycA levels, but a decrease in IL-6.

Subsequent genetic analyses shed light into the complex nature of these associations including the important issue of causality. First, MR results showing a potentially causal effect of total n-3 PUFAs on higher circulating CRP and GlycA levels contrast the presumed anti-inflammatory effect of n-3 PUFAs, and argues against the widespread use of n-3 supplementation aimed at improving health by reducing inflammation-related disease risk. Second, our MR results confirm that total n-6 and LA increase GlycA (a pro-inflammatory marker), suggesting a pro-inflammatory effects of these FAs. The importance of n-6 FAs with regards to modulating inflammation, rather than n-3 FAs, was also highlighted by our MVMR analysis which provided evidence for a direct causal effect of total n-6 FAs increasing CRP and GlycA levels after accounting for the effect of total n-3 PUFA levels. In contrast, no independent effect on inflammatory marker levels was observed for total n-3 FAs.

Results from our further MR analyses using genetic variants within/nearby a key gene involved in fatty acid metabolism (FADS gene cluster), indicate that several PUFA-inflammatory marker associations may result from pleiotropic effects of FA-related genetic variants, rather than the effect of FA biosynthesis. These include the effect of LA on GlycA, effect of total n-3 PUFAs on CRP and GlycA, and effect of total n-6 PUFAs on GlycA. Our approach of investigating the effect of specific genes within the PUFA biosynthesis pathway is similar to recent work by Haycock et al49 who investigated the effect of FAD1 and FAD2 genes on cancer. They reported that genetically elevated PUFA desaturase activity, instrumented by one SNP (rs174546), was associated with higher risk of colorectal and lung cancer. Future research should investigate whether desaturase activity promotes certain cancer risk through increased inflammation.

N-3 and n-6 PUFAs act as competitive inhibitors disrupting each other’s desaturation and elongation36. Therefore, a greater concentration of n-6 PUFAs would result in a fewer number of endpoint n-3 PUFAs being synthesized. This could be one explanation for the lack of an anti-inflammatory effect of n-3 PUFAs in our work. This may also explain our finding of positive associations between total n-6:n-3 PUFA levels and all three inflammatory marker levels in our cohort analysis.

A higher n-6:n-3 ratio is thought to promote the pathogenesis of inflammation-related NCDs such as cardiovascular disease, arthritis and asthma50, and a meta-analysis of RCTs found that supplementation of a low n-6:n-3 PUFA ratio inducing diet decreased concentrations of inflammatory markers such as tumor necrosis factor-α and IL-651. Therefore, it is possible that n-6:n-3 PUFAs ratio is more important than concentrations of individual PUFAs with regards to the risk of inflammation-related NCDs. We were unable to conduct MR for n6:n3 ratio because of the lack of GWAS for this measure. This should be prioritized when such data becomes available in future.

We investigated the association between PUFAs and levels of CRP, IL-6 and GlycA, using two different methods and used sensitivity analyses to address potential biases that arise in epidemiological research. Despite this, we recognize several limitations. We used cross-sectional analyses which means it is difficult to infer causality and we cannot determine the temporal relationship between the exposure and the outcome or the directionality of this relationship. Additionally, the cohort analysis was limited to only participants reported as White and the GWAS used were only conducted in individuals of European descent. This limits the generalizability of the results to other populations. However, the cohort analysis provides useful context and provides evidence for associations between PUFA levels and biomarkers of inflammation, including GlycA, which is novel.

DHA and LA appear at the different ends of FA biosynthesis pathways (n-3 DHA is an endogenous product of a chain of desaturation and elongation reactions, while LA is a pre-cursor to these reactions on the n-6 PUFA pathway). This means that a comparison between the effects of these specific n-3 and n-6 PUFAs is difficult. Future large-scale, well-powered GWAS of specific PUFAs across the biosynthesis pathway at comparable stages are needed to help better understand this relationship.

There was sample overlap between the exposure and outcome GWAS data used for MR analyses, which may lead to overfitting. However, recent research suggests that the bias incurred through sample overlap is less substantial compared to biases produced by weak instruments or winner’s curse47. These biases are unlikely to be issues for our MR analyses as confirmed by various sensitivity analyses exploring these concerns specifically.

Heterogeneity in the two-sample MR analyses may have biased causal estimates towards the null. Although we did investigate the outcome of the heterogeneity within the GlycA analyses using MR-Clust, no clusters were specifically related to inflammation-related processes.

Lastly, although the FADS gene cluster encodes enzymes fundamental to PUFA biosynthesis suggesting that IVs within this region are more likely to satisfy MR assumptions, FADS variants have been shown to be highly pleiotropic52 and do not differentiate n-3 or n-6 effects. As such, our gene based analyses may capture pathways to inflammation through factors other than PUFAs, thus violating the exclusion-restriction assumption40.

Given the popular use of n-3 PUFAs as supplements to reduce the occurrence of inflammation-related NCDs27, 53, 54, our results are of public health importance. Importantly, our MR results suggest that n-6 PUFAs increase levels of CRP and GlycA when controlling for n-3 PUFAs and that a higher total n-6:n-3 ratio increases levels of all three biomarkers. As such, instead of focusing solely on n-3 PUFA consumption, public health messaging should move towards encouraging a lower total n-6 PUFA consumption and to achieve a healthy balance between n-3 and n-6 to help reduce the occurrence of inflammation-related NCDs.

However, as there is evidence that suggests that n-3 PUFA consumption is also beneficial to health because of its hypotriglyceridaemic effects55, a better understanding of the interplay between n-6 and n-3 effects, as well as potential beneficial non-inflammation related effects, is needed before considering implementing n-6 focussed changes to policy.

4. Conclusion

Our MR results suggest that higher LA and total n-6 PUFA levels increase GlycA, but not CRP or IL-6 levels. Our MR results also suggest that higher total n-3 PUFA levels increase CRP and GlycA levels, but do not increase IL-6 levels. We provide consistent evidence for positive associations between total n-6:n-3 ratio and levels of CRP, GlycA and IL-6, indicating that the ratio of total n-6:n-3 PUFAs may be more important with regards to systemic inflammation than individuals PUFAs. However, these results came from cross-sectional analyses of cohort data and we were unable to examine evidence of causality for the effects of total n-6:n-3 ratio using MR due to lack of GWAS data for total n-6:n-3 ratio. Nevertheless, our findings do not support a strong anti-inflammatory effect of n-3 PUFAs from both cohort and MR analyses. This finding requires further investigation given that n-3 PUFA supplementation widely used to reduce inflammation and risk of inflammation-related NCDs.

5. Methods

5.1 Cohort Analysis

5.1.1 Dataset and sample

The Avon Longitudinal Study of Parents and Children (ALSPAC) birth cohort recruited pregnant women resident in Avon, UK with expected dates of delivery between 1st April 1991 and 31st December 1992 56–58. There were 14, 203 unique mothers initially enrolled in the study and the initial number of pregnancies enrolled was 14,541 and 13,988 children were alive at 1 year of age. When the oldest children were approximately 7 years of age, the initial sample was bolstered with eligible cases who did not originally join the study. The total sample size for analyses using any data collected after the age of seven is therefore, 14,833 unique mothers and 15,447 pregnancies which resulted in 15,658 foetuses. Of these, 14,901 were alive at 1 year of age. The offspring, their mothers and the mother’s partners are regularly followed up. Additional information on ALSPAC is presented in the supplement eMethods. For eligibility criteria see the study flowchart (eFigure 1) in the supplement.

For this study, ALSPAC participants were included if they had outcome (CRP, IL-6 and GlycA levels) and exposure (DHA, LA, total n-3 and total n-6) data at 24y. We excluded non-white participants (2% of study population) and this exclusion is unlikely to make any difference to the results due to low numbers of non-white individuals in the dataset. This gave a total of 2748 individuals (36.10% males) eligible for our analysis.

5.1.2 PUFAs and inflammatory marker assays

Participants fasted overnight, or >6 hours if being seen in the afternoon, before attending the clinic for blood sampling at 24 years old. Blood samples were centrifuged immediately to isolate plasma and stored at -80 °C. There were no freeze-thaw cycles during storage.

Total n-3, n-6, DHA and LA PUFA levels and plasma GlycA levels were measured using a high-throughput proton (1H) Nuclear Magnetic Resonance (NMR) metabolomics platform (Nightingale, UK)42, 59. We also created a n-6:n-3 ratio (total n-6 PUFAs divided by total n-3 PUFAs). HsCRP was measured by automated particle-enhanced immunoturbidimetric assay (Roche UK, UK). IL-6 was measured by enzyme-linked immunosorbent assay (OLINK, UK).

The samples were collected when the participansts were 24 years old. Full dtails regarding sample processing, NMR analysis, and data processing have been provided elsewhere60–62. GlycA levels ranged from 0.84 to 2.25 mmol/L, and IL-6 levels ranged from 1.82 to 10.60 Normalized Protein eXpression (NPX) Log2 scale. CRP had detection limits of 0.15-80mg/L and measures outside this were removed. CRP levels ranged from 0.15 to 70.05 mg/L.

5.1.3 Assessment of Covariates

We used maternal self-reported highest educational qualification and highest occupation of either parent (measures of social economic position (SEP)), maternal and paternal smoking pattern during pregnancy (measures of pregnancy health), participants’ sex, and participants’ smoking status and drinking status at age 24 as covariates. A detailed description of how these variables were collected/coded is provided in the supplement eMethods.

5.1.4 Statistical analysis

Multivariable linear regression analyses were used to examine the cross-sectional associations of PUFA levels with GlycA, CRP and IL-6 levels at age 24y. Regression models were adjusted for potential confounders. CRP and IL-6 levels were not normally distributed and were log-transformed. To help with the interpretation of results, we z-transformed both exposure and outcome data, and so the effect estimates represent the increase in outcome in standard deviation (SD) per SD increase in exposure. Primary analyses used the whole sample, and in secondary analyses we stratified by sex.

We used multiple imputation (MI) to impute missing exposure, outcome and covariate data in the eligible sample (N=2748), described in detail in the supplement eMethods.

5.2 Two-sample Mendelian Randomisation Analysis

5.2.1 Study design and data sources

Genome-wide significant single nucleotide polymorphisms (SNPs) (P< 5·0 x10-8) were selected as IVs for MR analyses. The SNPs were harmonized, aligning the genetic association for exposure and outcome on the effect allele using the effect allele frequency (see eMethods for details). Following harmonisation of the PUFA SNP instrument with the corresponding outcome data (inflammatory biomarker), palindromic SNPs were removed and the remaining SNPs underwent LD clumping. Steiger filtering was then applied, meaning SNPs were removed if they explained more variance in the outcome than in the exposure. The number of SNPs removed due to being palindromic, in linkage disequilibrium (LD) or as a result of Steiger filtering for each analysis is reported in eTable 1. There were 46-58 SNPs available to investigate the effect of PUFAs on GlycA levels, 17-39 SNPs available to investigate the effect of PUFAs on CRP levels and 11-44 SNPs available to investigate the effect of PUFAs on IL-6 levels. Harmonized SNP information for each relationship are presented in eTables 2-13. See eMethods for further details on instrument selection.

5.2.2 Statistical analysis

For MR, the inverse variance weighted (IVW)63 method was used as a primary analysis to calculate effect estimates. MR-Egger, weighted median, and weighted mode methods were used as sensitivity analyses as these methods make different assumptions about instrument validity. Findings were considered robust if results were consistent across primary and sensitivity analyses.

We report F-statistics for IVs, as a measure of instrument strength. Presence of heterogeneity between the SNP effect estimates was assessed using Cochran’s Q test in the IVW analysis and the Rucker’s Q test in the MR-Egger analyses. We used the MR-Egger intercept and MR Pleiotropy Residual Sum and Outlier global test (MR-PRESSO) to investigate the presence of pleiotropy. See eTable 14 for a description of MR methods and sensitivity analyses used64.

5.2.3 MR-Lap analysis to investigate bias from sample overlap, weak instrument and Winner’s Curse

The fatty acid GWAS, the CRP GWAS and GlycA GWAS were conducted using UKB data. This sample overlap may modify biases caused by the use of weak instruments and winner’s curse. To overcome this, the MR-Lap method was used to assess potential bias in observed effect estimates due to sample overlap, winner’s curse, or weak instruments47. Greater detail of MR-Lap method is provided in eMethods.

5.2.4 Additional MR analysis focusing on specific PUFA genes

MR was conducted using genetic instruments (SNPs) from within or close to (+/-500 kb) the FADS gene cluster (FADS1, FADS2 and FADS3; chromosome 11: 6115601452–6116591523) and ELOVL2 gene (chromosome 6: 1019801992–1110441624) from the DHA GWAS and the LA GWAS. SNPs were selected and harmonised using the same method as in the primary analysis and details are presented in the eMethods. The number of SNPs available for each analysis is presented in eMethods.

5.2.5 MR-Clust analysis to explore heterogeneity in MR instruments

SNPs can influence the outcome in distinct ways (e.g., via distinct biological mechanisms)48. This can lead to high levels of heterogeneity between individual SNP-outcome/SNP-exposure causal estimates within an instrument. Therefore, we decided to use MR-Clust48 to investigate whether individual causal estimates fell into distinct clusters based on effect magnitude. We used this method when evidence of association was consistent between cohort and MR analyses but there was strong evidence of heterogeneity in the latter. Functional annotations of the genomic loci associated with identified clusters was obtained from FUMAGWAS (https://fuma.ctglab.nl). See further details in eMethods.

5.2.6 Multivariable Mendelian Randomisation analysis

MVMR65 is an extension of MR that estimates the direct effect of each exposure on the outcome, rather than the overall total effect of the exposures (eFigure 2 of the supplement). We conducted a MVMR analysis to estimate the individual direct causal effects of total n-3 and n-6 PUFAs on the biomarkers of inflammation independently of each other. See further details in eMethods.

5.3 Software

Cohort analyses were performed using STATA version 17.0. All other analyses were performed in R Software version 4.1.0.

Contributions

Daisy C.P. Crick: Conceptualization, Methodology, Investigation, Writing – original draft, Writing – review & editing. Hannah Jones: Writing – Conceptualizaion, Supervision, Methodology, review & editing. Sarah Halligan: Supervision, Writing – review & editing. Golam M. Khandaker: Conceptualizaion, Methodology, Supervision, Writing – review & editing. George Davey Smith: Methodology, Writing – review & editing.

Correspondence to Daisy Crick

Competing interests

Authors declare no relevant potential conflict of interest related to the material presented in the article.

Data availability

Data needed to evaluate the conclusions presented in this paper are provided in the manuscript and/or the supplementary material. Additionaly ALSPAC data can be requested from the ALSPAC executive committee and reasonable requests from bona fide researchers. GWAS data is publicly available using the OpenGWAS website (https://gwas.mrcieu.ac.uk)

Code availability

Code for data management and statistical analysis has been made available in daisycrick (github.com).

Funding

The UK Medical Research Council and Wellcome (Grant ref.: 217065/Z/19/Z) and the University of Bristol provide core support for ALSPAC. This publication is the work of the authors and DC will serve as guarantor for the contents of this paper. A comprehensive list of grants funding is available on the ALSPAC website (http://www.bristol.ac.uk/alspac/external/documents/grant-acknowledgements.pdf); This research was specifically funded by Wellcome Trust and MRC (core) (Grant ref.: 76467/Z/05/Z), MRC (Grant ref.: MR/L022206/1) and Wellcome Trust (Grant ref.: 8426812/Z/07/Z).

This work was supported in part by the GW4 BIOMED DTP (D.C., MR/N0137941/), awarded to the Universities of Bath, Bristol, Cardiff and Exeter from the Medical Research Council (MRC)/UKRI.

GDS, GMK and HJ work within the MRC Integrative Epidemiology Unit at the University of Bristol, which is supported by the Medical Research Council (MC_UU_00011/1).

GMK acknowledges funding support from the Wellcome Trust (grant no: 201486/Z/16/Z and 201486/B/16/Z), the UK Medical Research Council (grant no: MC_UU_00032/06; MR/W014416/1; and MR/S037675/1), and the UK National Institute of Health Research Bristol Biomedical Research Centre (grant no: NIHR 203315).

HJ is supported by the NIHR Biomedical Research Centre at University Hospitals Bristol and Weston NHS Foundation Trust and the University of Bristol. The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care.

Acknowledgements

This work was carried out using the computational facilities of the Advanced Computing Research Centre, University of Bristol - http://www.bristol.ac.uk/acrc/.

We are extremely grateful to all the families who took part in this study, the midwives for their help in recruiting them, and the whole ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists, and nurses.

Footnotes

  • ↵* Joint senior authors

References

  1. 1.↵
    Videla LA, Hernandez-Rodas MC, Metherel AH, Valenzuela R. Influence of the nutritional status and oxidative stress in the desaturation and elongation of n-3 and n-6 polyunsaturated fatty acids: Impact on non-alcoholic fatty liver disease. Prostaglandins, Leukotrienes and Essential Fatty Acids 2022; 181: 102441.
    OpenUrl
  2. 2.↵
    Collaborators GRF. Global burden of 87 risk factors in 204 countries and territories, 1990-2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet 2020; 396: 1223–49.
    OpenUrlCrossRefPubMed
  3. 3.↵
    Calder PC, Albers R, Antoine JM, et al. Inflammatory Disease Processes and Interactions with Nutrition. British Journal of Nutrition 2009; 101: 1–45.
    OpenUrlPubMedWeb of Science
  4. 4.
    Calabrò P, Golia E, Yeh ET. CRP and the risk of atherosclerotic events. Seminars in Immunopathology; 2009: Springer; 2009. p. 79–94.
    OpenUrl
  5. 5.
    Calabro P, Golia E, TH Yeh E. Role of C-reactive protein in acute myocardial infarction and stroke: possible therapeutic approaches. Current Pharmaceutical Biotechnology 2012; 13: 4–16.
    OpenUrlCrossRefPubMed
  6. 6.
    Sabaawy HE, Ryan BM, Khiabanian H, Pine SR. JAK/STAT of all trades: linking inflammation with cancer development, tumor progression and therapy resistance. Carcinogenesis 2021; 42: 1411–9.
    OpenUrl
  7. 7.
    Rothwell PM, Wilson M, Price JF, Belch JF, Meade TW, Mehta Z. Effect of daily aspirin on risk of cancer metastasis: a study of incident cancers during randomised controlled trials. The Lancet 2012; 379: 1591–601.
    OpenUrl
  8. 8.
    Rothwell PM, Fowkes FGR, Belch JF, Ogawa H, Warlow CP, Meade TW. Effect of daily aspirin on long-term risk of death due to cancer: analysis of individual patient data from randomised trials. The Lancet 2011; 377: 31–41.
    OpenUrl
  9. 9.
    Connelly MA, Otvos JD, Shalaurova I, Playford MP, Mehta NN. GlycA, a novel biomarker of systemic inflammation and cardiovascular disease risk. Journal of Translational Medicine 2017; 15: 219.
    OpenUrl
  10. 10.
    Dowlati Y, Herrmann N, Swardfager W, et al. A meta-analysis of cytokines in major depression. Biol Psychiatry 2010; 67: 446–57.
    OpenUrlCrossRefPubMedWeb of Science
  11. 11.
    Fries GR, Walss-Bass C, Bauer ME, Teixeira AL. Revisiting inflammation in bipolar disorder. Pharmacology Biochemistry and Behavior 2019; 177: 12–9.
    OpenUrl
  12. 12.↵
    Oliveira J, Oliveira-Maia A, Tamouza R, Brown A, Leboyer M. Infectious and immunogenetic factors in bipolar disorder. Acta Psychiatrica Scandinavica 2017; 136: 409–23.
    OpenUrl
  13. 13.↵
    Currenti W, Godos J, Alanazi AM, et al. Dietary Fats and Cardio-Metabolic Outcomes in a Cohort of Italian Adults. Nutrients 2022; 14: 4294.
    OpenUrl
  14. 14.↵
    El-Badry AM, Graf R, Clavien PA. Omega 3 - Omega 6: What is right for the liver? J Hepatol 2007; 47: 718–25.
    OpenUrlCrossRefPubMedWeb of Science
  15. 15.↵
    Miles EA, Childs CE, Calder PC. Long-Chain Polyunsaturated Fatty Acids (LCPUFAs) and the Developing Immune System: A Narrative Review. Nutrients 2021; 13.
  16. 16.↵
    Sampath H, Ntambi JM. Polyunsaturated fatty acid regulation of gene expression. Nutr Rev 2004; 62: 333–9.
    OpenUrlCrossRefPubMedWeb of Science
  17. 17.↵
    Sokoła-Wysoczańska E, Wysoczański T, Wagner J, et al. Polyunsaturated Fatty Acids and Their Potential Therapeutic Role in Cardiovascular System Disorders-A Review. Nutrients 2018; 10.
  18. 18.↵
    Burns-Whitmore B, Froyen E, Heskey C, Parker T, San Pablo G. Alpha-Linolenic and Linoleic Fatty Acids in the Vegan Diet: Do They Require Dietary Reference Intake/Adequate Intake Special Consideration? Nutrients 2019; 11.
  19. 19.↵
    Hodge L, Salome CM, Peat JK, Haby MM, Xuan W, Woolcock AJ. Consumption of oily fish and childhood asthma risk. Medical Journal of Australia 1996; 164: 137–40.
    OpenUrlPubMedWeb of Science
  20. 20.↵
    Calder PC. n-3 polyunsaturated fatty acids, inflammation, and inflammatory diseases. Am J Clin Nutr 2006; 83: 1505s–19s.
    OpenUrlPubMed
  21. 21.↵
    Chaddha A, Eagle KA. Omega-3 fatty acids and heart health. Circulation 2015; 132: e350–e2.
    OpenUrlFREE Full Text
  22. 22.↵
    Kris-Etherton PM, Fleming JA. Emerging nutrition science on fatty acids and cardiovascular disease: nutritionists’ perspectives. Adv Nutr 2015; 6: 326s–37s.
    OpenUrl
  23. 23.↵
    Pischon T, Hankinson SE, Hotamisligil GKS, Rifai N, Willett WC, Rimm EB. Habitual Dietary Intake of n-3 and n-6 Fatty Acids in Relation to Inflammatory Markers Among US Men and Women. Circulation 2003; 108: 155–60.
    OpenUrlAbstract/FREE Full Text
  24. 24.
    Lopez-Garcia E, Schulze MB, Meigs JB, et al. Consumption of trans fatty acids is related to plasma biomarkers of inflammation and endothelial dysfunction. The Journal of Nutrition 2005; 135: 562–6.
    OpenUrlAbstract/FREE Full Text
  25. 25.
    Kiecolt-Glaser JK, Belury MA, Andridge R, Malarkey WB, Glaser R. Omega-3 supplementation lowers inflammation and anxiety in medical students: a randomized controlled trial. Brain, Behavior, and Immunity 2011; 25: 1725–34.
    OpenUrl
  26. 26.↵
    Hu FB, Manson JE. Omega-3 Fatty Acids and Secondary Prevention of Cardiovascular Disease—Is It Just a Fish Tale?: Comment on “Efficacy of Omega-3 Fatty Acid Supplements (Eicosapentaenoic Acid and Docosahexaenoic Acid) in the Secondary Prevention of Cardiovascular Disease”. Archives of Internal Medicine 2012; 172: 694–6.
    OpenUrlCrossRefPubMedWeb of Science
  27. 27.↵
    Goldberg RJ, Katz J. A meta-analysis of the analgesic effects of omega-3 polyunsaturated fatty acid supplementation for inflammatory joint pain. Pain 2007; 129: 210–23.
    OpenUrlCrossRefPubMedWeb of Science
  28. 28.↵
    MacLean CH, Mojica WA, Morton SC, et al. Effects of omega-3 fatty acids on lipids and glycemic control in type II diabetes and the metabolic syndrome and on inflammatory bowel disease, rheumatoid arthritis, renal disease, systemic lupus erythematosus, and osteoporosis: Summary. AHRQ Evidence Report Summaries. Rockville (MD): Agency for Healthcare Research and Quality (US); 2004.
  29. 29.↵
    Kelley DS, Taylor PC, Nelson GJ, Mackey BE. Arachidonic acid supplementation enhances synthesis of eicosanoids without suppressing immune functions in young healthy men. Lipids 1998; 33: 125–30.
    OpenUrlCrossRefPubMedWeb of Science
  30. 30.
    Thies F, Miles EA, Nebe-von-Caron G, et al. Influence of dietary supplementation with long-chain n− 3 or n− 6 polyunsaturated fatty acids on blood inflammatory cell populations and functions and on plasma soluble adhesion molecules in healthy adults. Lipids 2001; 36: 1183–93.
    OpenUrlCrossRefPubMedWeb of Science
  31. 31.↵
    Kakutani S, Ishikura Y, Tateishi N, et al. Supplementation of arachidonic acid-enriched oil increases arachidonic acid contents in plasma phospholipids, but does not increase their metabolites and clinical parameters in Japanese healthy elderly individuals: a randomized controlled study. Lipids in Health and Disease 2011; 10: 1–12.
    OpenUrl
  32. 32.↵
    Shen S, Gong C, Jin K, Zhou L, Xiao Y, Ma L. Omega-3 Fatty Acid Supplementation and Coronary Heart Disease Risks: A Meta-Analysis of Randomized Controlled Clinical Trials. Front Nutr 2022; 9: 809311.
    OpenUrl
  33. 33.↵
    Gracious BL, Chirieac MC, Costescu S, Finucane TL, Youngstrom EA, Hibbeln JR. Randomized, placebo-controlled trial of flax oil in pediatric bipolar disorder. Bipolar Disorders 2010; 12: 142–54.
    OpenUrlCrossRefPubMedWeb of Science
  34. 34.↵
    Sarris J, Mischoulon D, Schweitzer I. Omega-3 for bipolar disorder: meta-analyses of use in mania and bipolar depression. J Clin Psychiatry 2012; 73: 81–6.
    OpenUrlCrossRefPubMed
  35. 35.↵
    Bazinet RP, Layé S. Polyunsaturated fatty acids and their metabolites in brain function and disease. Nature Reviews Neuroscience 2014; 15: 771–85.
    OpenUrlCrossRefPubMed
  36. 36.↵
    Spiller GA. Handbook of Lipids in Human Nutrition. Boca Raton (FL): CRC press; 1995.
  37. 37.
    Wei Y, Meng Y, Li N, Wang Q, Chen L. The effects of low-ratio n-6/n-3 PUFA on biomarkers of inflammation: a systematic review and meta-analysis. Food Funct 2021; 12: 30–40.
    OpenUrlCrossRefPubMed
  38. 38.↵
    Innes JK, Calder PC. Omega-6 fatty acids and inflammation. Prostaglandins, Leukotrienes and Essential Fatty Acids 2018; 132: 41–8.
    OpenUrlCrossRefPubMed
  39. 39.↵
    Ferrucci L, Cherubini A, Bandinelli S, et al. Relationship of plasma polyunsaturated fatty acids to circulating inflammatory markers. The Journal of Clinical Endocrinology & Metabolism 2006; 91: 439–46.
    OpenUrl
  40. 40.↵
    Sanderson E, Glymour MM, Holmes MV, et al. Mendelian randomization. Nature Reviews Methods Primers 2022; 2: 1–21.
    OpenUrl
  41. 41.↵
    Würtz P, Kangas AJ, Soininen P, Lawlor DA, Davey Smith G, Ala-Korpela M. Quantitative Serum Nuclear Magnetic Resonance Metabolomics in Large-Scale Epidemiology: A Primer on -Omic Technologies. American journal of epidemiology 2017; 186: 1084–96.
    OpenUrlCrossRefPubMed
  42. 42.↵
    Soininen P, Kangas AJ, Würtz P, Suna T, Ala-Korpela M. Quantitative serum nuclear magnetic resonance metabolomics in cardiovascular epidemiology and genetics. Circulation: cardiovascular genetics 2015; 8: 192–206.
    OpenUrlAbstract/FREE Full Text
  43. 43.↵
    Borges MC, Haycock PC, Zheng J, et al. Role of circulating polyunsaturated fatty acids on cardiovascular diseases risk: analysis using Mendelian randomization and fatty acid genetic association data from over 114,000 UK Biobank participants. BMC Medicine 2022; 20.
  44. 44.↵
    Clayton GL, Borges MC, Lawlor DA. From menarche to menopause: the impact of reproductive factors on the metabolic profile of over 65,000 women. medRxiv 2022: doi:10.1101/2022.04.17.22273947
    OpenUrlAbstract/FREE Full Text
  45. 45.↵
    Ligthart S, Vaez A, Võsa U, et al. Genome analyses of> 200,000 individuals identify 58 loci for chronic inflammation and highlight pathways that link inflammation and complex disorders. The American Journal of Human Genetics 2018; 103: 691–706.
    OpenUrlCrossRefPubMed
  46. 46.↵
    Ahluwalia TS, Prins BP, Abdollahi M, et al. Genome-wide association study of circulating interleukin 6 levels identifies novel loci. Hum Mol Genet 2021; 30: 393–409.
    OpenUrl
  47. 47.↵
    Mounier N. Bias correction for inverse variance weighting Mendelian randomization. Genetic Epidemiology 2023; 47: 314–31.
    OpenUrlCrossRef
  48. 48.↵
    Foley CN, Mason AM, Kirk PDW, Burgess S. MR-Clust: clustering of genetic variants in Mendelian randomization with similar causal estimates. Bioinformatics 2020; 37: 531–41.
    OpenUrl
  49. 49.↵
    Haycock PC, Borges MC, Burrows K, et al. The association between genetically elevated polyunsaturated fatty acids and risk of cancer. EBioMedicine 2023; 91: 104510.
  50. 50.↵
    Simopoulos AP, Cleland LGCE. Omega-6/omega-3 essential fatty acid ratio: The Scientific Evidence. Scandinavian Journal of Nutrition 2004; 48: 49–50.
    OpenUrl
  51. 51.↵
    Wei Y, Meng Y, Li N, Wang Q, Chen L. The effects of low-ratio n-6/n-3 PUFA on biomarkers of inflammation: a systematic review and meta-analysis. Food & Function 2021; 12: 30–40.
    OpenUrl
  52. 52.↵
    Chen Y, Lu T, Pettersson-Kymmer U, et al. Genomic atlas of the plasma metabolome prioritizes metabolites implicated in human diseases. Nature Genetics 2023; 55: 44–53.
    OpenUrlCrossRef
  53. 53.↵
    Simopoulos AP. Omega-3 fatty acids in inflammation and autoimmune diseases. J Am Coll Nutr 2002; 21: 495–505.
    OpenUrlCrossRefPubMedWeb of Science
  54. 54.↵
    Bouzan C, Cohen JT, Connor WE, et al. A quantitative analysis of fish consumption and stroke risk. Am J Prev Med 2005; 29: 347–52.
    OpenUrlCrossRefPubMed
  55. 55.↵
    Siriwardhana N, Kalupahana NS, Moustaid-Moussa N. Health benefits of n-3 polyunsaturated fatty acids: eicosapentaenoic acid and docosahexaenoic acid. Adv Food Nutr Res 2012; 65: 211–22.
    OpenUrlCrossRefPubMed
  56. 56.↵
    Boyd A, Golding J, Macleod J, et al. Cohort Profile: the ‘children of the 90s’--the index offspring of the Avon Longitudinal Study of Parents and Children. Int J Epidemiol 2013; 42: 111–27.
    OpenUrlCrossRefPubMedWeb of Science
  57. 57.
    Fraser A, Macdonald-Wallis C, Tilling K, et al. Cohort Profile: the Avon Longitudinal Study of Parents and Children: ALSPAC mothers cohort. Int J Epidemiol 2013; 42: 97–110.
    OpenUrlCrossRefPubMedWeb of Science
  58. 58.↵
    Northstone K, Lewcock M, Groom A, et al. The Avon Longitudinal Study of Parents and Children (ALSPAC): an update on the enrolled sample of index children in 2019. Wellcome Open Res 2019; 4: 51.
    OpenUrl
  59. 59.↵
    Soininen P, Kangas AJ, Würtz P, et al. High-throughput serum NMR metabonomics for cost-effective holistic studies on systemic metabolism. Analyst 2009; 134: 1781–5.
    OpenUrlCrossRefPubMed
  60. 60.↵
    Soininen P, Kangas AJ, Wurtz P, et al. High-throughput serum NMR metabonomics for cost-effective holistic studies on systemic metabolism. Analyst 2009; 134: 1781–5.
    OpenUrlCrossRefPubMed
  61. 61.
    Inouye M, Kettunen J, Soininen P, et al. Metabonomic, transcriptomic, and genomic variation of a population cohort. Mol Syst Biol 2010; 6: 441.
    OpenUrlAbstract/FREE Full Text
  62. 62.↵
    Goulding N, Goudswaard L, Hughes D, et al. Inflammation proteomics datasets in the ALSPAC cohort [version 1; peer review: 1 approved, 1 approved with reservations]. Wellcome Open Research 2022; 7.
  63. 63.↵
    Lawlor DA, Harbord RM, Sterne JA, Timpson N, Davey Smith G. Mendelian randomization: using genes as instruments for making causal inferences in epidemiology. Statistics in Medicine 2008; 27: 1133–63.
    OpenUrlCrossRefPubMed
  64. 64.↵
    Crick D, Sanderson E, Jones H, et al. Glycoprotein Acetyls and Depression: testing for directionality and potential causality using longitudinal data and Mendelian randomization analyses. medRxiv; 2022.
  65. 65.↵
    Sanderson E, Davey Smith G, Windmeijer F, Bowden J. An examination of multivariable Mendelian randomization in the single-sample and two-sample summary data settings. International Journal of Epidemiology 2019; 48: 713–27.
    OpenUrlCrossRefPubMed
Back to top
Posted November 28, 2023.
Download PDF

Supplementary Material

Data/Code
Email
Relationship between Polyunsaturated Fatty Acids and Inflammation: evidence from cohort and Mendelian randomization analyses
Daisy C. P. Crick, Sarah Halligan, George Davey Smith, Golam M. Khandaker, Hannah J. Jones
medRxiv 2023.11.27.23299099; doi: https://doi.org/10.1101/2023.11.27.23299099
Twitter logo Facebook logo LinkedIn logo Mendeley logo
Citation Tools

Subject Area