Educational note: addressing special cases of bias that frequently occur in perinatal epidemiology

Abstract

The epidemiologic study of pregnancy and birth outcomes may be hindered by several unique and challenging issues. Pregnancy is a time-limited period in which severe cohort attrition takes place between conception and birth and adverse outcomes are complex and multi-factorial. Biases span those familiar to epidemiologists: selection, confounding and information biases. Specific challenges include conditioning on potential intermediates, how to treat race/ethnicity, and influential windows of prolonged, seasonal and potentially time-varying exposures. Researchers studying perinatal outcomes should be cognizant of the potential pitfalls due to these factors and address their implications with respect to formulating questions of interest, choice of an appropriate analysis approach and interpretations of findings given assumptions. In this article, we catalogue some of the more important potential sources of bias in perinatal epidemiology that have more recently gained attention in the literature, provide the epidemiologic context behind each issue and propose practices for dealing with each issue to the extent possible.

Associations between Residential Proximity to Oil and Gas Drilling and Term Birth Weight and Small-for-Gestational-Age Infants in Texas: A Difference-in-Differences Analysis

“…By using term birth weight and SGA as infant health end points, our approach incorporates gestational age and gestation length as components of these outcomes. This minimizes the potential for bias but also means we cannot interpret these results with respect to gestational age or gestation length (Neophytou et al 2020;Wilcox 2001). Second, many other components of drilling infrastructure (e.g., pipelines, compressor stations, retention ponds) (U.S. Department of Energy 2009) may have been present in our study area that could have produced air pollution that affects infant health that are beyond the scope of this project.…”

Low and High Ambient Temperatures during Pregnancy and Birth Weight among 624,940 Singleton Term Births in Israel (2010–2014): An Investigation of Potential Windows of Susceptibility

“…Similarly, we fitted generalized linear models with binomial family and logit link function to estimate odds ratios (ORs) and 95% CIs for tLBW. To avoid potential bias due to the use of separate models for each trimester, we modeled temperatures during all three trimesters in a single model ( Neophytou et al. 2021 ; Wilson et al.…”

“…We extracted information on 714,599 live-born infants with estimated last menstrual period (LMP) dates from 1 January 2010 (when exposure data first become available) through 18 March 2014. Because our data set did not include births after 2014, we excluded births with an estimated LMP after 18 March 2014 (n = 7,543) to avoid fixed cohort bias (Neophytou et al 2021;Strand et al 2011a). After births without information on birth weight (n = 105) or birth address (n = 13,180) were excluded, there were 693,771 births.…”

Methodological Considerations for Preterm Birth Research

“…Perinatal study populations are dynamic and complex because the reproductive process spans from fertilization and implantation to clinically recognized pregnancy, and further to birth and early childhood. Processes of selection (e.g., implantation failure, early pregnancy losses) and attrition (e.g., stillbirths, neonatal deaths) render these populations incompletely observable ( 18 , 19 ). Studies investigating outcomes of pregnancy are subject to further selection, because recruitment is often based on a convenient sample of women willing and able to access routine care services early in pregnancy.…”

“…For example, only including live born preterm infants, as often happens in such analyses, essentially adjusts for pregnancy loss. However, if an unmeasured confounder is associated with both pregnancy loss and preterm birth ( Figure 2A ), a non-causal pathway will be created potentially biasing estimates ( 19 ). Selection bias can also be induced through inappropriate variable treatment.…”