Simulations and directed acyclic graphs explained why assortative mating biases the prenatal negative control design

Area of Science:

• Epidemiology
• Biostatistics
• Genetics

Background:

• Negative control designs are valuable for assessing prenatal exposure-outcome associations via in utero mechanisms.
• Assortative mating can introduce bias into negative control designs, but the mechanisms remain unclear.

Purpose of the Study:

• To elucidate how mutual adjustment of maternal and paternal exposure in regression models can effectively account for assortative mating.
• To explain the underlying reasons for bias introduction by assortative mating in these designs.

Main Methods:

• Utilized directed acyclic graphs to illustrate potential biases when maternal and paternal effects are modeled separately.
• Conducted a simulation study to empirically test the impact of increasing assortative mating on effect estimates.
• Compared bias in effect estimates from models with and without mutual adjustment strategies.

Main Results:

• Models lacking mutual adjustment exhibited increased bias with higher levels of assortative mating.
• Maternal and paternal effect estimates were mutually biased, diminishing the observed difference from the true effect.
• Mutually adjusted models demonstrated no such bias, preserving the integrity of effect estimates.

Conclusions:

• Mutual adjustment for maternal and paternal exposure effectively mitigates bias stemming from assortative mating in negative control designs.
• This approach enhances the reliability of conclusions drawn from prenatal exposure-outcome association studies.
• Further considerations regarding limitations of mutual adjustment are discussed.