Assessing the Impact of Exposure Misclassification in Case-Control Studies of Self-Reported Medication Use.

Area of Science:

• Obstetrics and Gynecology
• Pediatrics
• Pharmacoepidemiology

Background:

• Recall bias can significantly impact observational studies on prenatal medication use.
• Accurate assessment of medication exposure during pregnancy is crucial for understanding birth defect risks.

Purpose of the Study:

• To evaluate the effect of exposure misclassification on previous studies linking NSAID use in early pregnancy to amniotic band syndrome (ABS).
• To assess potential bias in studies examining nonsteroidal anti-inflammatory drugs (NSAIDs) and ABS risk.

Main Methods:

• Utilized data from the National Birth Defects Prevention Study (NBDPS) (1997-2011).
• Included 189 infants with ABS and 11,829 controls.
• Employed probabilistic bias analysis with external validation parameters, conducting multidimensional analyses to address parameter transportability uncertainty.

Main Results:

• Misclassification-adjusted estimates showed attenuated confounding-adjusted estimates when higher specificity in cases or sensitivity in controls was assumed.
• In specific scenarios (higher case sensitivity, Sp ≥ 0.9), adjusted estimates indicated upward bias.

Conclusions:

• Bias magnitude is dependent on misclassification extent and mechanism.
• External validation parameters were not directly transferable; direct application may be limited.
• In the absence of robust validation, bias analysis and simulation studies are essential for interpreting medication effects on pregnancy outcomes.