Invited commentary: the perils of birth weight--a lesson from directed acyclic graphs

Area of Science:

• Epidemiology
• Biostatistics
• Public Health

Background:

• The association between birth weight and infant mortality is well-established.
• A paradox exists where low birth weight infants in high-risk populations sometimes exhibit lower mortality than those in low-risk populations.
• This phenomenon, termed the "birth weight paradox," challenges conventional understanding.

Purpose of the Study:

• To investigate the underlying causes of the birth weight paradox.
• To critically evaluate the impact of statistical adjustments on observational study results.
• To elucidate the role of confounding and bias in epidemiological research.

Main Methods:

• Utilized directed acyclic graphs (DAGs) to model causal relationships.
• Analyzed the impact of adjusting for intermediate variables (birth weight) in observational studies.
• Applied principles of causal inference to epidemiological data.

Main Results:

• The birth weight paradox arises from bias introduced by adjusting for birth weight, a variable influenced by exposure and sharing common causes with mortality.
• Directed acyclic graphs visually demonstrate how such adjustments can distort true associations.
• The paradox serves as a clear illustration of Simpson's Paradox in epidemiological contexts.

Conclusions:

• Statistical adjustments for variables affected by exposure and related to outcome can introduce significant bias.
• Directed acyclic graphs offer a robust framework for identifying and understanding sources of bias in epidemiological studies.
• The causal role of birth weight in infant mortality requires careful consideration, avoiding biased analytical approaches.