Introduction/Background: Metformin is the first-line therapy for type 2 diabetes (T2D); however, a considerable inter-individual variability in glycemic response is observed among patients. This heterogeneity suggests that metformin’s effects depend not only on drug exposure but also on the underlying metabolic and genetic factors. Methods: We applied a Gene–Environment interaction Mendelian Randomization (MR-G×E) in a cohort of 2743 individuals to investigate whether genetically influenced metabolite-HbA1c associations differ by metformin use. Metabolites associated with metformin response were used to establish metabolite-specific polygenic risk scores (PRSs) using metabolome-wide association study (mGWAS) variants. Generated PRS were used as genetic instruments within a one-sample, modified two-stage least squares model. An interaction term between PRS and metformin use was included to assess treatment-dependent genetic effects, adjusting for age, sex, body mass index, and genetic ancestry (principal components). Results: Metformin use significantly modified genetically influenced associations between 18 metabolites and HbA1c. Positive and negative PRS-metformin interaction effects indicated attenuation, strengthening or reversal of baseline genetic associations under treatment. Several amino acid metabolites, palmitoyl sphingomyelin (d18:1/16:0), and carbohydrate-related metabolite 1,5-anhydroglucitol showed specific patterns under metformin use. Interestingly, several metabolites (creatinine, gamma glutamylcitrulline, N-acetylthreonine, 3-methyl-2-oxovalerate, glycerol-3-phosphate, 1-(1-enyl-palmitoyl)-GPC (P-16:0), 1-(1-enyl-palmitoyl)-2-linoleoyl-GPC (P-16:0/18:2), sphingomyelin (d18:1/22:1, d18:2/22:0, d16:1/24:1), fructose, and methyl-glucopyranoside (alpha + beta)) showed no basal causal association with HbA1c but exhibited significant interaction effect with metformin use, suggesting metabolic association only in the presence of metformin. Conclusions: These findings indicate that metformin modifies the genetically influenced metabolite-HbA1c relationships, exhibiting treatment-dependent metabolic effects that are not detectable with standard MR approaches. Incorporating pharmacological context into causal inference provides new insights into the metabolic basis for the variable metformin response and helps inform precision strategies for T2D management.

Introduction/Background: Observational links between circulating metabolites and lipid traits are often confounded by environmental factors. This study applied Mendelian randomization (MR) and gene–environment interaction MR (MR-G×E) to assess the average causal effects of genetically predicted metabolites on lipid traits and their modification by statin use. Methods: Participants (n = 2810) from Qatar Biobank were analyzed including genetics, metabolomics and lipid profiles. The cohort was divided into discovery (n = 1968) and validation (n = 842). One-sample MR using two-stage least squares regression was applied to estimate the average causal effect of metabolites on LDL, HDL, and Triglycerides, while MR-G×E analysis was also applied to investigate whether genetically predicted metabolite–LDL/HDL associations differ by statin use. Genetic variants associated with metabolites were used to generate a polygenic risk score (PRS). The interaction term between the PRS and statin use was incorporated in MR-G×E, while adjusting for confounders. Results: A total of 167 metabolites were commonly associated with LDL, HDL, and Triglycerides. One-sample MR showed no significant average causal effect of genetically predicted metabolites on lipid traits in the overall population. However, MR-G×E analysis revealed significant statin-dependent effects. Genetically predicted 2-amino-octanoate and cysteine–glutathione disulfide demonstrated strong and reproducible positive interaction effects with LDL levels in statin users in both discovery and validation cohorts. While no overall average causal effects were observed, MR-G×E revealed that genetically predicted metabolites can influence LDL levels in the context of statin use. Conclusions: These findings highlight the importance of incorporating pharmacological exposures in MR analyses and support the role of drug–metabolite interactions in precision lipid management.