Why match? Investigating matched case-control study designs with causal effect estimation.

Area of Science:

• Epidemiology
• Biostatistics
• Public Health

Background:

• Matched case-control designs are common in public health for confounding control.
• Traditional analysis uses conditional logistic regression, yielding conditional, not causal, odds ratios.
• Matching's primary benefit is often considered efficiency, not causal inference.

Purpose of the Study:

• To investigate case-control weighted targeted maximum likelihood estimation (ccTMLE) for marginal causal effects in matched designs.
• To compare ccTMLE in matched versus unmatched designs for information yield on marginal causal effects.
• To determine optimal study designs for causal effect estimation in public health.

Main Methods:

• Employed case-control weighted targeted maximum likelihood estimation (ccTMLE).
• Compared ccTMLE performance in matched and unmatched case-control study designs.
• Utilized procedures requiring knowledge of prevalence probabilities (van der Laan, 2008).

Main Results:

• ccTMLE can estimate marginal causal effects in matched case-control studies.
• Unmatched designs may provide more information for estimating marginal causal effects compared to matched designs.
• The choice of design impacts the efficiency and interpretability of causal effect estimates.

Conclusions:

• While matching offers efficiency, unmatched designs might be preferable for estimating marginal causal effects using ccTMLE.
• Researchers interested in causal effects may benefit more from unmatched designs in many public health scenarios.
• Further exploration of ccTMLE in different epidemiological designs is warranted.