Is light drinking during pregnancy okay? Experts respond

UPDATED 4PM: New research drawing on data from the UK’s Millennium Cohort Study, a large study tracking the long term health of children born in the UK, suggests that light drinking during pregnancy does not harm a child’s behavioural or intellectual development.

woman drinkingThe research, published today in the Journal of Epidemiology and Community Health (A BMJ journal) drew on a sample of 11,513 children and interviewed participants’ mothers about their drinking patterns during pregnancy and other social and economic factors that may impact a child’s development, when the children were aged 9 months.

Of the mothers interviewed, around 6% said they never consume alcohol, while a further 60% chose to abstain during pregnancy. Just under 26% said they were light drinkers (defined as 1-2 units of alcohol per week or at any one time), while one in 20 (5.5%) were moderate drinkers (3-6 units per week) and 2.5% were heavy or binge drinkers during their pregnancy (7+ units per week or 6 at one sitting).

Children whose mothers were heavy drinkers were more likely to be hyperactive, and have behavioural and emotional problems than children whose mothers chose not to drink during pregnancy.

But there was no evidence to suggest that the behavioural or intellectual development of children whose mums were light drinkers during the pregnancy had been compromised. On the contrary, the study’s results appear to show that children born to light drinkers were 30% less likely to have behavioural problems than children whose mothers did not drink during pregnancy.

The researchers further found that: “After taking account of a wide range of influential factors, these children [of light drinkers] achieved higher cognitive scores than those whose mums had abstained from alcohol while pregnant.”

Current Ministry of Health and Alcohol Advisory Council recommendations in New Zealand advise women should abstain from drinking alcohol during pregnancy. ALAC states:

“Even 1 or 2 drinks can affect your baby’s learning. Drinking more than that increases the risks of greater damage. Although this damage does not always happen when a mother drinks (which explains why some pregnant women have had the odd drink without apparent harm to their baby) it is impossible to know when harm will occur.”

Registered journalists can log into the SMC Resource Library or contact the SMC for a copy of the paper.

In light of the findings of the JECH paper, the Science Media Centre asked experts in the field for their reaction to the research (below). The Australian SMC has also rounded up comments.


Dr Pat Tuohy, Chief Advisor – Child & Youth Health, Population Health Directorate, Ministry of Health comments:

“I have major criticisms of the conclusions in the attached press release drawn from this paper. The Authors themself were a lot more circumspect in their conclusions. Basically it confirms that if there are any adverse effects from drinking small amounts of alcohol during pregnancy they are subtle and not easy to find in a study like this. The paper only applies to women of European ethnicity, as other ethnic groups were excluded from analysis. It is subject to a range of biases and confounding factors. It does not show that drinking during pregnancy is better for your baby.


“1 There is clearly a confounding effect of socioeconomic status and maternal capability, self efficacy etc (however measured) on these results. i.e. brighter, better off, more capable mothers have similarly capable children. This is shown by the trend towards statistical non-significance as these confounding factors are entered into the statistical model. (Models A-E generally show a decreasing OR as you enter more confounders, indicating that the confounders are likely to be the major explanation for the findings). The study provides no explanation of why the ORs for teetotalers are so much higher on many scores. Surely this points to an unrecognised confounder.

“2 The actual Odds Ratios for the SDQ scores fail to reach significance in model E (The best corrected for confounders) except for the increase in emotional symptoms in the children of heavy/binge drinkers, which is not a surprising finding. All the other ORs include 1, which means that they are non-significant findings.

“3 Similarly the Z score confidence intervals in Table 3 all include zero (indicating a non significant effect) except for the light drinkers group for naming vocab, and picture similarities. However the CIs for these two criteria are both 0.01 to 0.14, so while statistically significant are only of marginal significance. However it is clear that the outcomes are not worse for these outcomes.

“My response on reading the paper is that the widely different numbers in the different exposure groups are going to give different weightings to the confidence intervals. This means that with only 284 mothers in the heavy/binge drinkers group the power of the study to show a real effect of alcohol using these tests is small, but with almost 3000 in the light drinking group it will be easier to show a real effect. The fact that a behavioural effect was shown in the children of heavy/binge drinkers on the SDQ is therefore quite significant, and probably meaningful. An indicator of the lack of power of this study to detect significant clinical differences is the lack of adverse outcomes in the IQ scores of children of heavy/binge drinkers – we would have expected to see that finding.

“The choice of the SDQ as the mental and emotional health test is a problem, because although it has a high-ish concordance with DSM IV assessments (about 0.75) it is a screening tool, not diagnostic. It is also not as sensitive or specific if a single informant (parent only) is used, as was the case here. The mental health and behavioural consequences of mild FASD are subtle and may not be identified well by the SDQ. The researchers recognised that parental ethnicity is an potential confounding factors but chose to exclude motehrs of ‘non-european’ ethnicity instead of attempting to correct for this confounder. – not clear why. Also ‘higher risk’ mothers are excluded from the analysis, as they were less likely to participate in the 5 year old sweep.

Our advice:

“We continue to advise that no alcohol in pregnancy is the best option. This study will however provide some reassurance to mothers who drank lightly in early pregnancy, and will reinforce that it is never too late to stop.”

Dr Helen Moriarty, senior lecturer and addiction medicine practitioner, primary health care and general practice, University of Otago, Wellington comments:

“On the face of it this interesting paper would seem to challenge existing advice. However, there are flaws which cast doubt on the findings. Firstly it was a retrospective survey. Women were first asked about their drinking during pregnancy 9 months after the pregnancy had ended.

“Secondly there are numerical issues that the authors failed to address: the initial (9 month) survey was of 18,552 women which was an 85% response rate therefore the real bottom line is 21,825 ( true denominator). That 15% would have refused to participate at all immediately raises a query since the drinking habits of these refusers is not known and the outcome on their children was not measured.

“Further problems with numbers appear in table 1 where data is reported for only 11,513 women (less than 53% of the 21,825). Subsequent results on behaviour outcomes etc only account for 11,374 (52% of 21,825). There were issues with missing data on the key outcome measures, ie incomplete data sets for all boys and girls. I think enough said from me. It will be a good paper to use to teach students about critical appraisal!”

Dr Sheryl Parackal, research fellow, School of Population Health, University of Auckland comments:

“1. The results observed in this study is limited to the outcomes measured in the study and hence in no way can imply that there are “NO effects” for light or any drinking in pregnancy.

“2. The result that children born to light drinkers are better off than children born to women “who stopped drinking in pregnancy” is not surprising because the majority of those who stopped in pregnancy are likely to have done so after “recognizing pregnancy”, which could be any time from 4 weeks to 8 weeks or even later.

“The findings of our NZ study (Alcohol in pregnancy study 2005) indicates ~ half of women who were pregnant at the time of the study had drunk alcohol before realising they were pregnant and a high proportion of these women had binged in this period. For most women, pregnancy starts when they realise they are pregnant and hence don’t report drinking prior to this unless probed/asked. So the reference group used for the comparisons may not be appropriate.”

Dr Rosemary Marks, developmental paediatrician at Auckland’s Starship Hospital and President of the Paediatric Society of New Zealand comments:

“The cohort members’ infants in this study were aged 9 months when the questions were asked about mothers drinking during pregnancy. One of the crucial times in relation to alcohol and pregnancy is the first few weeks after conception. You are asking people to recall what they were drinking 18 months ago when they may not have been aware they were pregnant. It is a difficulty which plagues all studies on this.

“Most women are well aware these days that there is concern about drinking in pregnancy, so there is an incentive to show yourself in a good light, a tendency to minimise the amount of drinking reported. That is a potential confounder, getting truthful information about how much people actually did drink.

“There was a very clear relationship between smoking and the drinking categories. Women who never drank actually had quite a high rate of smoking. Women who were light drinkers had the lowest rate of smoking. In the analysis, smoking was grouped with a number of other variables. Could the small differences between the groups have been explained by smoking rather than drinking?

“When you use ‘not in pregnancy’ as a reference point, you may have a problem there. Does ‘not in pregnancy’ mean anytime from conception onwards or does it mean ‘not once you knew you were pregnant’?

“The advice that has always been given is that we do not know whether any drinking in pregnancy is safe or not and therefore until we have that information the advice is not to drink. The study doesn’t contradict the advice given, it says it is probably not harmful so pregnant mothers who have the odd glass of wine can be slightly reassured that they are not doing their children harm.”

Dr Trecia Wouldes, senior lecturer, psychological medicine at School of Medicine, University of Auckland comments:

“Kelly and colleagues have published further findings from the large Millenium Cohort Study that suggest children born to mothers defined as “light” drinkers were not at increased risk for clinically relevant behavioural or early learning difficulties.

“However, as with their earlier study of 3-year-old children, the findings from this follow-up study of 5-year-olds should be interpreted with caution as findings in both studies also show that “light” drinkers were more socially and economically advantaged than non-drinkers. This means children of “light drinkers” may have had better antenatal and postnatal care and live in environments where parents are better educated and there are more resources to support optimal cognitive and behavioural development.”

“In addition, as the authors have pointed out there are a number of methodological limitations to their study that make it difficult to interpret these results. First, because of the way they have defined “light” drinking this category may have included a heterogeneous group of mothers who may have had 1 or 2 drinks during their entire pregnancy and mothers who consumed 1 or 2 drinks per week throughout their pregnancy.

“In order to have any confidence in the level of alcohol exposure to children in the “light” group, measures of absolute alcohol and the timing across trimesters of maternal consumption is needed. In addition, maternal reports of alcohol consumption were obtained at 9 months after birth requiring the mother to reconstruct from memory her drinking behaviour.

“Second, the developmental outcomes these researchers have reported were all based on questionnaires or reports provided by the mother. Independent evaluations of these children by preschool teachers or observations from developmental specialists using more in depth and comprehensive measures of behaviour and learning may have strengthened or negated their findings.

Finally, although epidemiological studies can provide important information about maternal drinking patterns in women that is representative of the larger population, they usually do not have the resources to investigate in detail the environmental context of that drinking or the timing and frequency of alcohol, tobacco and other drug use during the pregnancy. Indeed the authors have noted in their discussion of the findings (p. 7), that given the impact of social advantage found in this and other studies of mothers who also reported light drinking, it is likely that “…rather than the direct physiochemical nature of the intrauterine environment it is likely that social circumstances to a large part are responsible for the relatively low rates of subsequent behavioural difficulties and the cognitive advantage in children whose mothers were “light” drinkers.”

Professor Jennie Connor, Head of Department, Preventative and Social Medicine, Dunedin School of Medicine, University of Otago comments:

“The evidence of “safety” as well as the evidence of “harm” that comes from this study may be affected by a number of shortcomings of the cohort study design in answering questions about cause and effect. There is no way to completely eliminate the effects of other lifestyle factors, which are associated with different drinking patterns, on the outcomes for the children.

“Women who drink at a low level during pregnancy have other characteristics that make their children at less risk of the “problems” being studied, and women who drink heavily during pregnancy are very likely to have other characteristics that put their children at risk of these problems. The investigators have adjusted the results for some of these factors but cannot adjust for them all.

“When the results that are adjusted for all of the factors that they have data on (Model E in Table 2), the findings are not statistically significant. This means that it is plausible that there is no difference between the children of light drinkers and non-drinkers in terms of behavioural problems. The authors acknowledge all of these issues.

“I consider that the claim made in the press release that “Children born to light drinkers were 30% less likely to have behavioural problems than children whose mothers did not drink during pregnancy” is not supported by the data presented in the paper. Once the comparisons are adjusted for some of the other differences between the mothers, the difference is not 30% and not statistically significant. The authors do not appear to make this claim in the paper.

“As well as this source of bias in the results (confounding), alcohol studies require drinking data to be self reported, and in this study data were collected 9 months after the child’s birth. Alcohol consumption is almost always under reported, and this measurement error is exacerbated when recall over long periods is required. So if the findings were true, there is no way of knowing what the true “safe” dose is.

“These problems do not mean the study was poorly carried out; they are inherent in the design. There are well known examples of results from cohort studies being found to be wrong when a randomised trial has been subsequently conducted on the same question. In this case the definitive study would require randomly allocating pregnant women to different levels of drinking during pregnancy and measuring the effects on the children, which would be clearly unethical.

“It is biologically unlikely that a moderate dose of alcohol would improve brain function in a fetus and a heavier dose would impair it, since alcohol is well known to be a neurotoxin.

“The authors say in the last paragraph “However, causal inference based on observational data is limited, and further work to tease out aetiological relationships is needed. This means that we cannot draw the conclusion that drinking in pregnancy is safe from this study.

“The most important message of the study is: Children of mothers who drank 7 or more units per week (one bottle of wine over the week) or 6 units at one sitting were more likely to be hyperactive, and have behavioural and emotional problems than children whose mothers chose not to drink during pregnancy.

“This is the most important message because we must take a precautionary approach to this issue. There will be individual variation in the effects of alcohol on the fetus and/or lack of clarity about the dose and or timing of alcohol exposure that can cause harm. There is no risk associated with abstinence.”

Dr Susan Morton, principal investigator, Growing Up in New Zealand longitudinal Study, University of Auckland comments:

“The Millenium Cohort Study is an important observational population longitudinal study that is looking at how children develop in the UK in the 21st century. However it did collect its first set of data from mothers when their children were at least 9 months of age. Hence the information collected about maternal drinking in pregnancy is subject to recall bias (because it is collected after the time when the drinking occurred and after the child had been born).

“It seems that there was also only one question asked of mothers about their frequency of drinking in pregnancy and mothers were not asked about when and how much they may have drunk at different stages of their pregnancy, particularly in the first trimester as compared to later in the pregnancy.

“This is especially important for planned and unplanned pregnancy (approximately 50% pregnancies reported to be unplanned) where heavy drinking at pre-pregnancy levels may continue into the first critical weeks of fetal development before a pregnancy is recognized. Further the assignment of ” never drinkers” in the MCS seems to rely on reported information from mothers only after their children were born rather than reflecting the situation before they were pregnant, which may have introduced further bais if mothers habits had changed after the birth of their children.

“In New Zealand currently mothers are advised to abstain from alcohol during pregnancy because (as Dr Connor points out) of the potentially harmful effects of heavy alcohol consumption. What is not well understood is what effects if any (on child development) lighter or more moderate drinking might have on children’s outcomes at birth and beyond. Because of this lack of scientific evidence about cause and effect we err on the side of caution and recommend that women abstain completely – as we know this is “safe”.

“Clearly conducting a scientific experiment to ascertain whether any level of drinking in pregnancy is safe or not would be unethical, therefore we do rely on observational, self-reported data to help us understand this issue further.

“To help to address this issue for our current New Zealand population the Growing Up in New Zealand longitudinal study has asked mothers and their partners about their drinking before and during their pregnancy at the first interview with over 7000 parents of the children in the new cohort. In contrast to the MCS we have asked the mothers and partners about their drinking when they were still pregnant, and have asked them to describe their drinking at each stage of their pregnancy (first trimester and beyond) as well as to compare this to their usual pre-pregnancy alcohol intake. The first results from these interviews are currently being prepared for release and will be available at the end of November (being released at parliament on November 25th).

“This is the first longitudinal study in New Zealand that has recruited both mothers and partners during pregnancy (to follow over 7000 children from before birth until they are young adults) and that has recruited a cohort that are representative of all our current births (so include Maori, Pacific and Asian parents and children in sufficient numbers in addition to our European parents and children). The information we have collected in this new study will allow us to see what effect different frequency and timing of alcohol intake has on our children as they grow up in New Zealand today.”

“As well as collecting information about alcohol intake the Growing Up in New Zealand study has also collected a wealth of information about the parents, their families and the environments into which the children have been born. This will allow us to consider the relationship between drinking in pregnancy and child development in the context of all the other behaviours and influences that may confound (also contribute to) any association that is seen.

“This is extremely important as mothers who have different patterns of alcohol consumption are also likely to differ in other significant ways that may affect their child’s growth and development (for example we will also be able to concurrently consider mothers and partners smoking during pregnancy, mothers nutrition, parents education and other markers of family circumstances and deprivation during pregnancy and after the children are born).”