The Need for More Quasi‐experimental Studies of Alcohol Consumption During Pregnancy

The review by Gray, Mukherjee & Rutter [1] presents succinctly and accurately what we do and do not (unfortunately) know concerning the effects of alcohol consumption during pregnancy (ACDP). The review highlights two key problems related to extant human studies of ACDP: (i) the difficulties in understanding the implications of variability in the quantity, pattern and timing of alcohol exposure and (ii) the potential of correlated risks to confound the associations between ACDP and offspring neurodevelopmental traits. I will focus on the latter concern, although research obviously relies heavily on clearly defined and accurate measurement. One of the major recommendations the authors make is that researchers should use natural experiments, or quasi-experimental approaches, which help to delineate between the possible causal influence of ACDP and the effects of covarying risks. The purpose of this commentary is to emphasize this need further and provide additional examples of quasi-experimental designs to encourage additional research in this area. Interestingly, many of these approaches have been used to study the effects of an associated risk, that of maternal smoking during pregnancy (SDP) [2]. As Gray and colleagues [1] point out, the comparison of differentially exposed siblings is a research approach that can help to rule out confounding genetic and environmental risks [3,4]. Converging evidence from sibling comparison studies suggests that SDP is associated independently with pregnancy-related outcomes (e.g. birth weight and placenta ruptures) [5–8], consistent with a causal inference. However, recent sibling-comparison studies indicate that the association between SDP and later neurodevelopmental traits, such as intellectual abilities/academic achievement [7–10] and childhood conduct problems [6,7], are due to family background factors, not the teratogenic effects of SDP. As such, the findings suggest that the underlying causal mechanisms related to SDP depend upon the specific trait being studied. Very few sibling comparison studies, however, have explored ACDP [11]. The Children of Twins (CoT) design, an extension of the cousin comparison design that accounts for environmental factors that influence all family members in an extended family, and genetic factors passed down from the twin parent [12–14], can also be used to study prenatal risks. A CoT study found an independent association between SDP and offspring birth weight [13], providing converging evidence with sibling-comparison studies. There are two extant CoT studies of ACDP; they do not, however, provide consistent results concerning the association between ACDP and offspring attention deficit hyperactivity disorder [15,16]. Researchers are also incorporating numerous quasi-experimental approaches into studies to test simultaneously alternative causal processes and account for limitations/assumptions inherent in each. For example, a comparison of differentially exposed full siblings, half-siblings, full cousins (offspring of full siblings) and half-cousins (offspring of half-siblings) suggests that genetic factors passed down from mothers and fathers account for the statistical association between SDP and offspring academic functioning [10]. I am unaware of any such studies of ACDP. Many different quasi-experimental (or natural experiments) can be used to study ACDP (reviews in [2,17]). In designing future studies of ACDP, researchers should consider including multiple offspring per family, numerous mothers (and their offspring) in extended families, and twin families. Future studies should also consider including adopted individuals [18], partner drug use during pregnancy and biomarkers to use Mendelian randomization, as recommended by Gray et al.[1]. These designs will provide invaluable insight into the role of ACDP, but such research requires special resources (e.g. access to twin registries) and the passage of considerable time to follow a sample longitudinally. In the meantime, are there data sets available to researchers with multiple siblings, cousins or offspring of twins, which include assessments of ACDP and offspring adjustment? Such secondary data analysis projects frequently require the use of less than ideal assessments of alcohol use (e.g. [11]), but the field needs many studies, using multiple samples and designs (each with their own strengths and weaknesses) to draw stronger inferences regarding the putative effects of ACDP [17]. Certainly, more animal research and family studies of ACDP need to be completed. Given the potential role of confounding variables, however, the design and implementation of new natural experiments and the use of existing quasi-experimental studies will be crucial for our understanding of ACDP. Special thanks to Valerie Knopik and Amber Singh for their help preparing the text. The author is supported by grants from NICHD (HD056354 and HD053550) and Indiana University (Faculty Research Support Program).