Article Text

Download PDFPDF

Alcohol consumption in pregnancy as a risk factor for later mental health problems
  1. Kapil Sayal
  1. Section of Developmental Psychiatry, University of Nottingham, Nottingham NG7 2UH, UK

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

There has been increasing interest in recent years about the possible mental health risks for offspring as a result of prenatal alcohol exposure. In particular, there is considerable controversy about whether there is a safe threshold for alcohol consumption during pregnancy and whether international policy recommendations are based on evidence. This article briefly summarises the existing literature in relation to mental health outcomes in childhood, adolescence and early adulthood following prenatal alcohol exposure. It also highlights some of the possible pitfalls in the interpretation of observational epidemiological data.

There are currently considerable international differences in policy recommendations about alcohol use in pregnancy. Some countries such as the USA, Canada and France have recommended complete abstinence. This appears to be based on a precautionary approach, given that no clear safe threshold for drinking has been established. In contrast, until May 2007, guidelines in the UK were that pregnant women could safely drink one or two drinks per occasion up to 1–2 times per week (that is, up to a total of four drinks/units of alcohol per week). This approach appears to be based on the lack of available evidence demonstrating that low levels of alcohol consumption are associated with later adverse outcomes. These recommendations were recently revised to advise abstinence. However, this revision was based on the perceived greater clarity of a “no drinking” message rather than on the research evidence, and the guidelines remain ambiguous as they state that women who choose to drink can drink up to the previous limits.1


A recent systematic review carried out by the National Perinatal Epidemiology Unit focused on the literature relating to risks associated with low to moderate levels of alcohol consumption and binge drinking during pregnancy.2 There is more conclusive evidence related to higher levels of drinking, manifesting at the extreme end as fetal alcohol syndrome. In the USA, there has been acknowledgment of partial forms of these outcomes, and the prevalence of fetal alcohol spectrum disorder (FASD) had been estimated at around 1% of live births.3 More recently, in Europe, rates of FASD have been estimated at 2–4%.4 The majority of these children are described as having an alcohol-related neurodevelopmental disorder (ARND), manifesting in terms of problems with overactivity, inattention, behaviour or learning. These outcomes have been found to be associated with moderate levels of exposure (for example, an average of one drink per day) and these findings have been replicated in animal studies, suggesting that there may be critical vulnerable periods for neurodevelopment. There is greater controversy about the possible harmful effects of lower levels of drinking (ranging between occasional drinking and less than one drink per day during pregnancy). There has been limited published research investigating the role of low levels of exposure, which may reflect a publication bias if there have been mainly negative findings. Often where these findings have been presented in papers, the role of lower levels of alcohol use has been investigated as part of a wider range that includes higher levels of drinking. Therefore, findings involving lower levels have not usually been adjusted specifically for key confounding variables. However, at least two studies suggest that low levels of drinking during pregnancy may be associated with childhood mental health problems. A study in the USA found that up to 3–4 drinks per week was associated with behavioural problems at ages 6–7 years, after adjusting for confounders.5 These findings were recently extended by Sayal et al in the UK who found that less than one drink per week in the first trimester was associated with higher levels of mental health problems between the ages of 4–8 years as assessed by both parents and teachers.6 This association was only demonstrated in girls, highlighting the need for further replication of these findings.

The consumption of moderate levels (1–2 drinks per day) of alcohol during pregnancy has been found to be associated with childhood attentional and behavioural problems.711 There is increasing evidence that the risks from alcohol consumption in pregnancy may persist over time. Cohort studies that have reported data collected during adolescence and early adulthood have described ongoing learning problems, behavioural difficulties, and adult psychiatric disorders such as personality and substance use disorders.1214 However, these findings need to be interpreted cautiously in terms of the potential caveats outlined below.


Observational epidemiological research investigating the aetiological role of environmental factors during pregnancy in contributing to later psychopathology is beset by methodological difficulties. The key methodological concerns relating to bias and confounding will be described. First, bias can be considered in terms of selection and information bias. Case-control studies are particularly susceptible to selection bias because of their frequent reliance on clinical samples to identify the cases. These are subject to referral bias and may provide misleading estimates of the risk from the exposure. Ideally, to demonstrate causal associations involving prenatal alcohol exposure, prospective longitudinal studies involving large sample sizes which are representative of the wider population are required. However, there are inherent difficulties with this in longitudinal research. Recruitment approaches adopted by investigators (such as the levels of generalisability from the choice of baseline sample) and high levels of long-term loss to follow-up mean that findings need to be interpreted with caution. Sample attrition over time may also be systematically associated with key exposure, confounder and outcome variables meaning that findings are only partially generalisable. International differences and changes in temporal trends in drinking during pregnancy also mean that findings need to be interpreted cautiously. Pregnant women who consumed alcohol at a particular timepoint (such as during a particular decade) and in a particular culture may not be generalisable to a sample from a different time and place. For example, the correlates of women who drank heavily during pregnancy in the 1970s may be very different from those who drink heavily now, given the greater knowledge of the possible risks.

With regard to information bias, both exposure and outcome status are vulnerable to measurement errors. In particular, there may be recall bias with regard to ascertainment of the exposure status (especially if enquiries are made postnatally). Prospective research designs are required that ascertain the timing, frequency, amount, duration, pattern and variability of drinking across the course of the pregnancy. Even then, several studies suggest that an important risk period is before pregnancy recognition,12 ,15 a period that is particularly vulnerable to recall bias about the exposure. When relying on self-report, there is considerable risk of underestimating alcohol consumption. This can result in systematic misclassification with women who drink heavily reporting their consumption as moderate, and those who drink moderately reporting light or no consumption, thus contributing to misleading risk estimates. There may also be a tendency to provide socially desirable responses. Studies comparing levels of alcohol consumption elicited through concurrent and retrospective self-report have found that higher levels are reported retrospectively.16 Despite all these factors, self-reports of alcohol consumption have been shown to be valid in relation to predicting risk for neurobehavioural outcomes.11 Methodological approaches that average out weekly amounts of drinking might mask binge episodes of alcohol consumption and fetal exposure to peak levels of alcohol. Similarly, the averaging out of levels over longer time periods might obscure risks related to particular timing of the exposure. Finally, outcome measures should ideally involve a range of informants during childhood (for example, parents and teachers) or involve reliable and valid diagnostic interview measures.

Second, there are a range of alternative factors that might explain an apparent relation between alcohol consumption in pregnancy and later mental health difficulties. Prenatal confounding variables including the use of other substances, smoking, background socioeconomic factors, maternal anxiety and depression, and maternal nutrition, might be associated with both the exposure and the outcome. As above, the measurement of confounders (in particular, the use of other substances and smoking during pregnancy) should also be rigorous to minimise information bias. Poorly measured or unmeasured confounders may otherwise account for a spurious association between a risk factor and an outcome. Even if analyses control for a large number of confounding variables, residual confounding remains a possibility. The postnatal environment, particularly involving parental mental health or quality of parenting, may include more proximal mechanisms through which the apparent risk is transmitted. Genetic predisposition, on both the maternal and paternal side, needs to be taken into account when considering how the risk might be transmitted. For example, an apparent association between prenatal alcohol exposure and childhood psychopathology disappeared when controlling for family risk for alcohol dependence.17 There are also likely to be gene-environment correlations whereby the propensity to drink during pregnancy may be closely related to the genetic risk of transmission of mental health disorders. Finally, given that not all children who have been exposed to high levels of alcohol consumption in pregnancy have adverse outcomes, there may be gene-environment interactions whereby a combination of risk factors is required.


There is a challenge for future research designs to account for the potential range of prenatal environmental risk factors for later mental health problems and how they might operate in combination with genetic risk. In the meantime, there is also a need for clarity from policy makers in terms of key public health and health education messages for pregnant women and women of child-bearing age who might be contemplating becoming pregnant. As yet, a safe threshold has not been conclusively established and it is important that accurate information about possible risks is available so that women can make informed decisions.



  • Competing interests: None.