Population stratification

Population stratification occurs when there exist population subgroups that experience both different phenotypic distributions and have different allele frequencies for genetic variants that might be used in MR. This can result in spurious (confounded) associations between genotype and phenotype (i.e., a trait that might be used as either the risk factor or outcome in MR). 

If genetic IVs used in an MR study were derived from GWAS in a population in which there were sub-groups (e.g., different ethnicities) that differed with respect to the distributions of both the genetic IVs and the phenotype (risk factor of interest in the MR study), then the genetic IV-risk factor association could be biased by confounding. If the genetic IV-outcome associations were determined in studies in which there were sub-groups with differing genetic and outcome distributions, the genetic IV-outcome association could be biased by confounding. Most GWASs and MR studies try to minimise this by including homogeneous groups of participants (e.g., of just one ethnicity) and/or by adjusting for genetic principal components that reflect different sub-groups within that population sample. 

References

• Lawlor DA, Harbord RM, Sterne JAC, Timpson NJ, Davey Smith G. Mendelian randomization: using genes as instruments for making causal inferences in epidemiology. Statistic in Medicine 2008;27:1133-1163.

Other terms in 'Sources of bias and limitations in MR':

• Assortative mating
• Canalization
• Collider
• Collider bias
• Confounding
• Dynastic effects
• Exclusion restriction assumption
• Harmonization failure (in two-sample MR)
• Homogeneity Assumption
• Horizontal Pleiotropy
• Independence assumption
• InSIDE assumption (in two-sample MR using aggregate data)
• Monotonicity assumption
• MR for testing critical or sensitive periods
• MR for testing developmental origins
• No effect modification assumption (Additional IV assumption)
• Non-linear effects
• Non-overlapping samples (in two-sample MR)
• Overfitting
• Pleiotropy
• Regression dilution bias (attenuation by errors)
• Relevance assumption
• Reverse causality
• Same underlying population (in two-sample MR)
• Statistical power/efficiency
• Vertical Pleiotropy
• Weak instrument bias
• Winner's curse