Causal Inference and Its Applications in Medical Research

The presentation titled “Causal Inference and Medical Research” explains why understanding causal relationships is crucial for clinicians, tracing the evolution from early theories about miasma to the modern microbiological view of infectious diseases.

It outlines five main topics: the motivation for studying causality, how epidemiology infers causation, recent advances in causal inference, and future directions.

Traditional epidemiological tools such as Mill’s methods (commonality, difference, co‑occurrence, dose‑response, and exclusion) and Koch’s postulates are described, highlighting their role in identifying disease risk factors.

Modern challenges in chronic non‑communicable diseases are discussed, emphasizing the complexity of multi‑factor causal chains and the need to distinguish host versus environmental contributors.

The article introduces cohort studies and randomized controlled trials (RCTs) as hierarchical evidence, illustrating how adjusting for confounders (e.g., smoking) can reveal true causal links between risk factors like obesity and mortality.

Advances in Mendelian randomization (MR) are detailed, showing how genetic variants serve as instrumental variables to infer causality, with examples of single‑sample two‑stage least squares and two‑sample Wald ratio methods, while noting limitations such as weak instruments and pleiotropy.

Future research directions include causal graph structure learning, integration of causal inference with differential equations, deep learning, and factor analysis, suggesting that improved causal tools will enhance disease treatment, including cancer therapies.