Mendelian randomization is a method that uses genetic variants as instrumental variables to infer causal relationships between risk factors and health outcomes, minimizing confounding and reverse causation. It leverages the random assortment of genes from parents to offspring, akin to a natural randomized controlled trial, providing more robust causal inferences in epidemiological studies.