Questions: Mendelian Randomization and Genetic Causal Inference

Question 1

A researcher proposes using a genetic variant in the serotonin transporter gene as an instrument to study whether depression causes cardiovascular disease. Later evidence shows the same variant also directly affects inflammatory pathways in the heart, independent of depression. Which instrumental variable assumption does this violate?

• A. Relevance — the variant does not strongly predict depression
• B. Independence — the variant associates with cardiovascular confounders
• C. Exclusion restriction — the variant affects the outcome through a pathway other than the exposure
• D. None — pleiotropic effects are permitted in Mendelian randomization

Answer: (C)

The exclusion restriction requires that the genetic instrument affects the outcome only through the exposure (depression), not through any independent pathway. A variant that directly influences cardiac inflammation violates this: it creates a back-door path from instrument to outcome that bypasses the exposure. This is the most common threat to MR validity — many genetic variants have pleiotropic effects, which is why sensitivity analyses like MR-Egger are used to test robustness.

Question 2

Why does Mendel's law of independent assortment make genetic variants useful as instrumental variables in epidemiology?

• A. Because genetic variants perfectly predict the exposure, eliminating measurement error
• B. Because alleles are assigned randomly at conception, creating natural balance across confounders like socioeconomic status and lifestyle
• C. Because genetic variants are fixed at birth and cannot be changed by behavior
• D. Because GWAS studies can identify variants with very large effect sizes on exposures

Answer: (B)

The power of MR as a causal method comes from randomization: just as an RCT randomly assigns treatment, Mendel's second law randomly assigns alleles during gamete formation. This random assignment is independent of confounders like diet, income, or smoking — factors that plague observational epidemiology. That a variant predicts exposure is necessary (relevance), but it's the *random* nature of allele inheritance that satisfies the independence assumption and gives MR its causal inference logic.

Question 3

The independence assumption in Mendelian randomization is analogous to randomization in a clinical trial, because alleles are randomly assigned at conception rather than based on an individual's lifestyle or social circumstances.

• T. True
• F. False

Answer: True

This is exactly right and is the core analogy that makes MR compelling. In an RCT, randomization ensures that treatment and control groups are balanced on observed and unobserved confounders. Mendel's law plays a comparable role: which allele you inherit at a locus is determined at fertilization by processes independent of your later environment, diet, or behavior. This is why MR is called 'nature's randomized controlled trial.' The analogy is imperfect — population stratification can violate independence — but the logic is sound.

Question 4

A genetic variant that strongly predicts heavy alcohol consumption is a valid Mendelian randomization instrument for studying alcohol's effect on liver disease, regardless of whether it also directly influences liver enzyme activity.

• T. True
• F. False

Answer: False

This describes a violation of the exclusion restriction. If the variant directly affects liver enzyme activity independent of alcohol consumption, it creates a pathway from the genetic instrument to the outcome that does not pass through the exposure. The exclusion restriction requires the instrument to affect the outcome *only through* the exposure. Strong prediction of the exposure (relevance) is necessary but not sufficient — all three IV assumptions must hold for a valid MR analysis.

Question 5

Explain why Mendelian randomization can estimate causal effects that observational epidemiology cannot, and identify the key assumption most likely to be violated.

Model answer: MR uses genetic variants as instruments to mimic randomization: because alleles are assigned randomly at conception and cannot be changed by behavior, the variant-exposure association is not confounded by lifestyle factors. By using only the variation in exposure predicted by the genotype, MR bypasses the confounding that makes observational associations unreliable. The assumption most likely to be violated is the exclusion restriction: genetic variants often have pleiotropic effects, influencing multiple biological pathways simultaneously. If the instrument affects the outcome through any pathway other than the target exposure, the causal estimate is biased.

Observational studies suffer because people who differ in exposure (e.g., alcohol consumption) also differ in many other ways (diet, stress, socioeconomic status) — these co-occurring factors confound the exposure-outcome association. MR exploits the biological lottery of inheritance: your genotype was assigned before you had a lifestyle. The exclusion restriction is the hardest assumption to verify because pleiotropy is biologically common, which is why modern two-sample MR includes sensitivity analyses (MR-Egger, weighted median) that can detect and partially correct for pleiotropic bias.