Mendelian randomization (MR) is a method that uses genetic variants to infer causal relationships between exposures and outcomes in epidemiological studies. It helps address confounding and reverse causation by leveraging genetic factors that are not influenced by environmental or observational biases. MR has been widely applied to various exposures and outcomes, including biomarkers, behaviors, and physiological measures. Recent developments include two-sample MR, bidirectional MR, network MR, and multiphenotype MR, which enhance the ability to infer causality. However, MR faces challenges such as pleiotropy, where a genetic variant affects multiple traits, potentially leading to biased results. To mitigate this, multiple genetic variants can be used to increase statistical power and test assumptions. Additionally, two-sample MR allows for the use of independent datasets to improve the reliability of causal estimates. Bidirectional MR helps distinguish between causal directions, while network MR explores complex interactions among multiple variables. Mediation analysis using MR can help identify intermediate factors in causal pathways. Factorial MR assesses the combined effects of multiple risk factors, and multiphenotype MR addresses the challenge of estimating causal effects when genetic variants are associated with multiple traits. MR also enables hypothesis-free inference by identifying potential associations from large-scale genetic data. Overall, MR provides a robust framework for causal inference in epidemiology, with ongoing developments aimed at improving its accuracy and applicability.Mendelian randomization (MR) is a method that uses genetic variants to infer causal relationships between exposures and outcomes in epidemiological studies. It helps address confounding and reverse causation by leveraging genetic factors that are not influenced by environmental or observational biases. MR has been widely applied to various exposures and outcomes, including biomarkers, behaviors, and physiological measures. Recent developments include two-sample MR, bidirectional MR, network MR, and multiphenotype MR, which enhance the ability to infer causality. However, MR faces challenges such as pleiotropy, where a genetic variant affects multiple traits, potentially leading to biased results. To mitigate this, multiple genetic variants can be used to increase statistical power and test assumptions. Additionally, two-sample MR allows for the use of independent datasets to improve the reliability of causal estimates. Bidirectional MR helps distinguish between causal directions, while network MR explores complex interactions among multiple variables. Mediation analysis using MR can help identify intermediate factors in causal pathways. Factorial MR assesses the combined effects of multiple risk factors, and multiphenotype MR addresses the challenge of estimating causal effects when genetic variants are associated with multiple traits. MR also enables hypothesis-free inference by identifying potential associations from large-scale genetic data. Overall, MR provides a robust framework for causal inference in epidemiology, with ongoing developments aimed at improving its accuracy and applicability.