Abstract
Ketogenic diets are popular among people aiming for weight management. Ketone supplementation has been linked to improved cognitive performance and increased risk of insulin resistance. We aim to identify common genetic variants that allow Mendelian randomization investigations into further potential effects of ketone metabolism. We set four premises that we believe any valid instrument for ketone metabolism should satisfy. These are: (1) location in a gene region relevant to ketone metabolism, (2) association with all three primary ketone bodies (acetone, acetoacetate, and beta-hydroxybutyrate), (3) no pleiotropic associations, (4) associations with positive control variables (cognitive performance, two-hour glucose, and insulin fold change). We considered gene regions containing variants previously associated with acetone. Four of these regions had biological relevance to ketone metabolism. Lead variants for three of these four regions (SLC2A4, HMGCS2, OXCT1) were associated with all three primary ketone bodies. One region (SLC2A4) was associated with two-hour glucose and insulin fold change; however, this region had strong pleiotropic associations with blood pressure. One region (OXCT1) showed an association with cognitive performance, and thus satisfied all our premises to be a valid instrument for ketone metabolism. In a complementary agnostic approach considering all genome-wide significant predictors of the three primary ketone bodies in turn, genetically predicted acetoacetate based on seven variants was associated with improved cognitive performance. However, several variants selected in this approach were not located in biologically relevant gene regions and were pleiotropic. Causal claims from Mendelian randomization will be most reliable when the instrumental variable assumptions are plausibly satisfied. We illustrate a framework to identify candidate instruments based on biological considerations.