MR Dictionary

An MR method that explores the separate and joint causal effects of two (or potentially more) risk factors on an outcome and is analogous to factorial randomized controlled trial (RCT) designs. Participants in a factorial MR study are categorised into different levels of each risk factor based on their genetic IVs. Where the genetic IV involves multiple SNPs, categories of high vs. low exposure might be defined by being above and below the median of a polygenic risk score (PRS) of those SNPs. For a genetic IV that includes a SNP, the categorisation will be one group of homozygotes versus a group of the heterozygotes and other homozygotes together, or three groups of homozygotes for one allele, heterozygotes, and homozygotes for the other allele. Regression analyses are then conducted with the genetic IV categories.

Genetic IVs for the two (or potentially more) risk factors are the categories defined by genetic variants. These categories need to fulfil the key IV assumptions. Factorial MR may have limited statistical power (in comparison to factorial RCTs and MR of single risk factors).

Factorial Mendelian randomization. Adapted with permission from Carter et al.  In this example, factorial MR is used to investigate the combined effects of BMI and alcohol on liver disease and liver disease biomarkers. Disease and biomarker levels are compared across the four categories of genetically instrumented combinations of low and high BMI and alcohol levels. Low and high BMI were defined as below or equal to and above the median of a weighted polygenic risk score (PRS) for BMI, respectively. For low and high alcohol categories, a single candidate genetic instrumental variable (ADH1B) was used as an instrument and homozygotes for the allele associated with higher alcohol consumption (high) were compared to those who were heterozygotes or homozygotes for the allele associated with lower alcohol consumption.
Figure 2.2 - Factorial Mendelian randomization. Adapted with permission from Carter et al. In this example, factorial MR is used to investigate the combined effects of BMI and alcohol on liver disease and liver disease biomarkers. Disease and biomarker levels are compared across the four categories of genetically instrumented combinations of low and high BMI and alcohol levels. Low and high BMI were defined as below or equal to and above the median of a weighted polygenic risk score (PRS) for BMI, respectively. For low and high alcohol categories, a single candidate genetic instrumental variable (ADH1B) was used as an instrument and homozygotes for the allele associated with higher alcohol consumption (high) were compared to those who were heterozygotes or homozygotes for the allele associated with lower alcohol consumption.

References

Other terms in 'Definition of MR and study designs':