Deconstructing the smoking-preeclampsia paradox through a counterfactual framework

Area of Science:

• Obstetrics and Gynecology
• Epidemiology
• Biostatistics

Background:

• Numerous studies report an inverse association between maternal smoking and preeclampsia, a paradoxical finding given smoking's known adverse pregnancy outcomes.
• This paradox may stem from biases in study design and analysis, rather than a true protective effect of smoking.

Purpose of the Study:

• To investigate the paradoxical association between maternal smoking during pregnancy and preeclampsia using a counterfactual framework.
• To explore and correct for selection bias and confounding factors contributing to this paradox.

Main Methods:

• A case-control study nested within the Icelandic Birth Registry (1309 women) was conducted.
• Directed acyclic graphs, logistic regression, mediation analysis, and sensitivity analyses were employed to identify and adjust for bias.
• Cases were defined as pregnancy-induced hypertension or preeclampsia after 20 weeks' gestation; controls were normotensive mothers.

Main Results:

• The biased analysis showed smoking reduced preeclampsia odds by 28% (OR 0.72), and 75% after stratification by gestational age (OR 0.25).
• Mediation analysis revealed the paradox's structure, indicating bias.
• The bias-adjusted odds ratio for smoking increasing preeclampsia risk was 1.22 (95% CI 0.41, 6.53).

Conclusions:

• The smoking-preeclampsia paradox is likely due to selection bias (prevalent cases at birth), unmeasured confounders, and controlling for a collider (gestation weeks).
• Adjusting for these biases suggests smoking may not be protective and could potentially increase preeclampsia risk.
• Future research must account for these methodological aspects when examining smoking and pregnancy outcomes.