Can non-responding mask or mimic drug effects on pregnancy outcome? Evaluation of case characteristics based on the national Embryotox cohort

Abstract

Women and their health care provider (HCP) often seek advice for drug safety in pregnancy at Teratology Information Services (TIS). In turn, TIS ask for details of drug exposure and pregnancy outcome. These data constitute a valuable basis for research on prenatal drug risks in many countries. Non-response to follow-up questionnaires, however, may cause biased study results. To assess the potential of non-response bias, this study based on the German Embryotox cohort compares maternal and HCP characteristics of responders and non-responders. Change in loss of follow-up rates over time is investigated using logistic regression. From 2010 until the end of 2020, 48,410 pregnant women and/or their HCP consented to participation in follow-up. Of these, 25.0 % did not return follow-up questionnaires. Loss rates were similar for patients and HCP but increased over time. Participants from semi-dense populated areas had a smaller loss rate (20.4 %) than those from rural (28.4 %) or urban areas (25.6 %). Responding women were older than non-responders, had a lower BMI, a more positive attitude towards pregnancy, a higher educational level, a lower number of previous pregnancies, smoked less, and indicated alcohol consumption more but social drugs less often. Non-response bias cannot be ruled out in studies based on observational data on drug use in pregnancy as those collected by TIS. However, differences between the complete and lost-to follow-up cohort do not suggest a particularly high or low risk profile for one of the cohorts that might substantially confound study results or even mask or mimic potential drug toxicity.