(P29-026-24) Developing an Objective Method to Improve the Causal Estimates of Dietary Traits on Metabolic Health Outcomes in Mendelian Randomization

Objectives: Mendelian randomization causal inference (MR) has clarified many relationships between dietary traits and health outcomes and is a cost and time-effective tool to vet causal relationships prior to performing randomized controlled trials. However, dietary traits are often associated with other behavioral traits and social determinants of health, misconstruing the true direct effects of the diet trait on the health outcome.

Methods: We are developing a method, Mendelian Randomization Selection of Lifestyle Associated Variants (MRS.LAVa), that objectively filters variants within the genetic instrument (GI). MRS.LAVa incorporates a phenome-wide association study (PheWAS) step including over 2,000 health and lifestyle-related traits within the UK Biobank to filter variants whose variant-trait association explains less variance than that of the trait of interest. We applied MRS.LAVa to dietary traits and health outcomes with established relationships, including fresh fruit on diastolic blood pressure (DBP) and oily fish on triglycerides (TG). The dietary data was obtained from a published GWAS of dietary traits from the UK Biobank, and the health outcomes were obtained from the MR-IEU Open GWAS Project. Statistical output from the unfiltered and filtered GI were compared using the inverse-variance weighted (IVW) method following standard harmonization steps.

Results: In the unfiltered oily fish-GI, intake of oily fish consumed weekly reduced TGs (27 variants; IVW β=-0.16, p=0.02, Cochrane’s Q=34.3), but statistical significance was attenuated after applying MRS. LAVa (7 variants; IVW β=0.01, p=0.95, Q=3.34). In contrast, pieces of fresh fruit consumed daily were not associated with DBP in the unfiltered method (43 variants; IVW β=-0.11, p=0.81, Q=264) but achieved statistical significance after the filtering (19 variants; IVW β=-1.04, p=0.035, Q=18).

Conclusions: Development of MRS.LAVa will continue, including refining the filtering method, benchmarking against existing methods that consider pleiotropy, testing other diet traits and metabolic outcomes, and making the tool publicly available. Improved MR methods will promote precision nutrition by identifying components of diet intake causally linked to aspects of metabolic health.

Funding Sources: This work was funded by the NIDDK and the NLM of the NIH.