Identification of a genetic risk factor for metformin-induced vitamin B12 deficiency

Abstract Aims/hypothesis Metformin, a mainstay of treatment for type 2 diabetes, can cause vitamin B 12 deficiency. Clinical risk factors have been identified but genetic factors remain undiscovered. Our objective was to identify and validate genetic predisposing factors and establish clinical utility. Methods Individuals with metformin-induced vitamin B 12 deficiency ( n =487) and metformin-tolerant control individuals ( n =6686) were identified in UK Biobank. Genome-wide association analysis was undertaken using logistic regression. Replication was undertaken in three cohorts: a Scottish cohort; the Diabetes Prevention Program Outcomes Study (DPPOS); and a separate cohort from Liverpool. In the Liverpool cohort, plasma metformin levels were also measured. Results Analysis identified a genome-wide significant non-synonymous SNP in the cubilin gene ( CUBN, rs1801222/p.S253F) associated with metformin-induced vitamin B 12 deficiency (additive model; adjusted p =1.86×10 −10 ; OR 1.56 [95% CI 1.36, 1.79] for AG vs GG genotype; OR 2.43 [95% CI 1.85, 3.20] for AA vs GG genotype), which was replicated in both the Scottish and the DPPOS cohorts. Vitamin B 12 deficiency occurred in 0.84–1.20% of non-metformin-exposed individuals regardless of rs1801222 genotype. However, a large interaction with metformin use was observed, with vitamin B 12 deficiency occurring at 6.02% in GG, 7.96% in GA and 12.84% in AA genotype groups. When followed up from metformin initiation, 10% with the AA genotype were vitamin B 12 deficient by 11 years vs 21 years for 10% of the GG group. Conclusions/interpretation The observed genetic association suggests that the rs180122 genotype should be considered a significant risk factor for metformin-induced vitamin B 12 deficiency. While clinical monitoring of serum vitamin B 12 levels in patients on metformin is inconsistently done, this finding highlights the potential clinical utility of targeted monitoring for certain subsets of individuals, including those genetically at high risk. Graphical Abstract