Causal knowledge as a prerequisite for confounding evaluation: an application to birth defects epidemiology

Area of Science:

• Epidemiology
• Biostatistics
• Reproductive Health

Background:

• Confounding variables complicate observational studies, particularly in understanding risks for birth defects.
• Statistical criteria are commonly used to identify confounders, but may not always reflect true causal relationships.
• Folic acid supplementation is a key area of research for preventing neural tube defects.

Purpose of the Study:

• To evaluate the role of statistical versus a priori causal knowledge in confounder identification.
• To determine the appropriate odds ratio for assessing the relationship between folic acid and neural tube defects.
• To demonstrate the utility of causal diagrams in epidemiological analysis.

Main Methods:

• A case-control study (Slone Epidemiology Unit Birth Defects Study, 1992-1997) was analyzed.
• Statistical criteria were applied to assess confounding.
• A priori causal knowledge, encoded using causal diagrams, was incorporated for comparison.
• Crude and adjusted odds ratios were calculated and compared.

Main Results:

• Statistical confounding adjustment yielded an odds ratio of 0.80 (95% CI: 0.62, 1.21).
• A priori causal knowledge indicated the adjusted odds ratio was invalid.
• The crude odds ratio of 0.65 (95% CI: 0.46, 0.94) was deemed appropriate based on causal reasoning.

Conclusions:

• Relying solely on statistical criteria for confounding can lead to invalid conclusions in epidemiological research.
• Integrating a priori causal knowledge, often visualized with causal diagrams, is essential for accurate risk assessment.
• The crude odds ratio suggests a protective effect of folic acid supplementation on neural tube defects, contrary to the adjusted finding.