Light Drinking and Pregnancy: A Review of the Research

[1]

Listed conclusion: “Women who regularly drink as little as two glasses of wine per drinking session while pregnant can adversely impact their child’s results at school.”

What the authors examined was how “light to moderate” drinking was associated with child school performance at age 11.  Drinking was categorized as never, <1 glass a week, 1-6 glasses a week, and 1+ glass per day.  They also categorized “binge” drinking as asking about how many times the person (they asked moms and dads) had consumed more than 2 pints of beer, 4 glasses of wine, or 4 pub measures of spirits in a single sitting during the first 3 months of pregnancy, with answers including never, 1-2 days, 3-4 days, 5-10 days, and 10+ days.  Academic outcomes were national curriculum tests.  In addition, some variables were considered for confounders (e.g., SES, marital status) but mental health was absent.

The researchers found no effect on cognitive performance of maternal alcohol use up to 1 drink per day (there was a linear increase in performance linked to paternal alcohol use, but that’s a topic for another day).  In fact, when the confounders were controlled for, more alcohol use was associated with greater performance at age 11 until 1+ drinks per day.  The immediate effect was for “binge” drinking (4 or more glasses of wine in a single sitting) with an immediate difference between the never and 1-4 days, even when the confounds are controlled for.  Notably, postnatal binge drinking was not associated with any changes in child cognitive performance.

Conclusion?  The results do not support the idea that light drinking is linked to lower cognitive performance.

Alcoholism: Clinical and Experimental Research, 2012[2].

Listed conclusion: “Reduced birth length and weight, microcephaly, smooth philtrum, and thin vermillion border are associated with specific gestational timing of prenatal alcohol exposure and are dose-related without evidence of a threshold. Women should continue to be advised to abstain from alcohol consumption from conception throughout pregnancy.”

In this study, the authors examined a large cohort of women who self-referred to a counseling program in California between 1978 and 2005.  Data on alcohol consumption was assessed during the pregnancy and the outcome variables of interest were physical deformations associated with FAS, which were assessed by a dysmorphologist who was blind to the women’s exposure.

The researchers assessed various confounds like smoking and exposure to other medications and drugs, maternal education, etc., yet none of them were included in the final analyses due to a very odd statistical decision.  The researchers decided to only include a variable as a confound if it altered the effect of alcohol by more than 15%.  I have never heard of such a decision and, in fact, confounds are supposed to be included en masse and based on both theoretical and statistical significance.  There are many valid reasons for why they need to be included and included in a certain manner that go beyond what I have time to explain here.  Suffice it to say that this is a problem and a bit of a big one.

Ignoring this problem, what did they find?  Well, they did find increased risk of certain physical features starting at 1 drink per day (7 drinks per week) and only for drinking in the second half of the first trimester.  There is no other information for risk less than this amount given the way they ran their analyses.  Thus, their statement that there is no evidence of a threshold is slightly disingenuous given they cannot assess the lower levels of alcohol consumption.  Also, 7 drinks per week, as you will see, would actually place someone in a “heavy” drinking group based on some of the more recent prospective studies being done.  This will be further discussed in the section on defining light drinking.

Conclusion?  The data as is does not support the idea that light drinking is associated with an increased risk of physical malformations, but nor does it reject the idea.

International Journal of Epidemiology, 2012[3].

Listed conclusion: “Even low amounts of alcohol consumption during early pregnancy increased the risk of spontaneous abortion substantially.”

In reality, this study is such a mess, I don’t quite know what to say about it, but will do my best (and luckily the researchers themselves acknowledge some of the problems here).  This was a large-scale Danish study examining pregnancy loss associated with alcohol exposure and included a huge cohort of women.  Prenatal alcohol exposure was assessed as women were referred to the study, so some reported prospectively and some retrospectively.  Notably, the restrospective assessment was from women who miscarried and then were referred.  Many confounds were considered and included and fetal death was examined in various time brackets (i.e., up to 12 weeks, 13-16 weeks, 17-21 weeks, and 22+ weeks).

When looking at the cohort as a whole, there were significant relationships between drinking and pregnancy loss starting with 2-3.5 drinks per week (which would definitely be classified as “light” by nearly all studies).  This only holds for miscarriage up to 16 weeks, after which alcohol has zero effect (even with higher levels of drinking).  The mess and problems come when the researchers analyzed the data only for those women who were referred early and before any pregnancy loss (i.e., the true prospective cohort).  With just these women, there was a statistically significant risk associated with .5 drinks per week up to 3.5 drinks per week of miscarriage in the first trimester (but not after), and no risk at all for any time period of 4 or more drinks per week.  So heavier drinking did not increase the risk, but lighter drinking did.  See what I mean by messy?

Now why exclude those who entered after a miscarriage?  Because, as discussed in this paper, other researchers have found that while alcohol is often underreported prospectively, it is often overreported after a pregnancy loss.  This means the inclusion of these individuals would bias results towards significance when in fact, they may be closer to the null.  And of course, the findings that there was no relationship at the highest level of drinking for those who reported prospectively raises huge questions about what exactly is being assessed here.  (Even more confusing, though not statistically significant, the odds-ratio estimate for fetal loss after the first trimester puts drinking as a protective factor, which makes no sense.)  The researchers also acknowledge that they don’t know the potential mechanism for their findings given that a large prospective study of 32,000 pregnancies involving risk of physical malformations  in relation to moderate alcohol consumption (which was classified as less than 6 drinks a day) found no effect of alcohol[4].

Conclusion?  Even the researchers admit this study has uncertainties to it, so it is not conclusive of anything really.

Alcohol Research and Health, 2011[5].

Listed Conclusion: “This study found that drinking at low to moderate levels during pregnancy is associated with miscarriage, stillbirth, preterm delivery, and sudden infant death syndrome (SIDS).”

First off, this isn’t a “study”, it’s a review of the research.  Second, it didn’t really find these things. The review was a lit review looking at these outcomes listed above, but also included an excellent section on methodological limitations in the research.  I would highly recommend reading this for an understanding of how hard this research can be.  But the question is what they found in their review, limitations notwithstanding.

Miscarriage?  They concluded that moderate to heavy use is definitely linked to miscarriage.  For light drinking, they report, “a review (Henderson et al., 2007) examining the impact of light to moderate prenatal alcohol exposure concluded that there is no consistent evidence for an increased risk of spontaneous abortion at these lower levels of exposure. The studies described above also seem to support that conclusion.” (p.88)

Stillbirth? Relationships were found for women drinking 14 or more drinks per week, more than 5 drinks per week, and ever drinking.  However, as noted in this review regarding the finding that ever drinking increased the risk: “Data were further analyzed to examine the amount of alcohol consumption, and the increased risk was almost completely attributed to those who consumed five or more drinks per week.” (p.88)

Preterm delivery?  The two studies that didn’t suck (which is basically how the authors put it – they were very unimpressed with the data in this category) found no effects of light to moderate drinking on preterm delivery.  One study found effects with 10 or more drinks per week, the other only found it for “binge drinking at any point in pregnancy and heavy drinking during the first trimester” (p.89)  Two other studies were discussed, but one didn’t measure level of alcohol consumption and the other only looks at days of drinking, not amount drank on a given day.

SIDS?  The authors report that there isn’t great data on this, with only small-scale studies on the matter that don’t include all confounds.  However, what we do have is mixed with some studies finding any alcohol early in pregnancy increases the risk, and some finding only binge drinking affects risk (which may or may not be due to assessment of drinking).  Interestingly, one study found decreases in infant alertness and increases in infant irritability along with lower sleep-related movements in women with the highest reported alcohol use, suggesting a possible mechanism for higher-levels of alcohol intake.

Conclusion?  Higher levels of alcohol use are related to all of these outcomes.  Light drinking, not so much, though there remains the possibility of a link for SIDS if large-scale research finds anything to support the smaller scale studies.

Alcohol, Health, and Research World 1997[6].

Listed conclusion: “Even a small amount of alcohol may affect child development.”

The link from NOFAS doesn’t work so I searched all articles from 1997 and one seems to be the one and that is what I will discuss here.  Again, this isn’t a study but a brief review of the literature and is focused predominantly on FAS and not the broader spectrum of disorders. This review highlights some important facts, such as how difficult it is to assess FAS and FASDs given the nature of development.  Cognitive deficits often do not show until later, at a time when the facial features associated with FAS are reduced due to normal changes in facial features (such as length and width).  As such, a lot of what is defined is done by using guesswork and assuming the same associations known to exist for more severe cases of FAS.  (I highlight this because it’s very interesting given that the neurocognitive and behavioural deficits are similar to those associated with a variety of other disorders and delays.  Similarly, FAS and FASDs often co-occur with other developmental delays and disorders.  There are many possibilities for this, including that alcohol is the underlying factor in many of these outcomes or that the other disorders and delays are responsible for the deficits attributed to prenatal alcohol exposure when FAS is not conclusively assigned.)

In terms of their review of dosage and effect, they state that it varies by outcome, with the effects on the central nervous system occurring at a threshold level (i.e., there is a safe amount of alcohol that one can drink before seeing any effects), but that the effects on postnatal growth are linear (i.e., there is a dose-response with even minimal drinking resulting in an effect).  In the research on this linear relationship[7], however, the researchers determined linearity by only looking at the jump from 0 drinks per day to 1 drink per day (i.e., 7 drinks per week).  They cannot ascertain if there is a threshold effect less than this amount.

Notably, the authors suggest that the research demonstrates that these growth deficits disappear in more advantaged groups.  Thus, there seems to be a prenatal alcohol exposure by SES interaction with more disadvantaged groups suffering the growth deficit at lower levels of alcohol exposure, but not more advantaged groups (e.g., [8]).  This again highlights the potential for other environmental factors to influence outcomes and interact with prenatal alcohol exposure.

 

Conclusion?  The data doesn’t say much of anything about “light” drinking, but supports the idea that there are difficulties associated with moderate-heavy drinking.  It also highlights factors in assessment, which is important.

Other Studies

Longitudinal Prospective Studies

There are two large-scale, longitudinal, prospective studies underway looking at the effects of alcohol use and later cognitive and behavioural outcomes.  I will discuss briefly how these studies have been conducted and what their findings are before turning to their flaws.

The first of these studies is out of the UK and uses the UK Millenium Cohort to determine the effects of light, moderate, or heavy drinking on postnatal outcomes.  Contrary to much of the aforementioned research which considered “light” drinking to be 1 drink per day, the definitions here were as follows: Light drinking was 1-2 units per week and occasion (i.e., if you drank 3 units once per week, you would not qualify), moderate drinking was 3-6 units per week or 3-5 units per occasion, and heavy drinking was 7 or more units per week or 6 or more units per occasion.  As you can see, the “light” category is significantly less than most other studies.  Of course, there was also an abstaining group.  In addition to assessing alcohol use and child outcomes, the researchers also controlled for various confounds including infant, socio-economic, and family psychosocial factors (including mother’s mental health, SES, etc.).  As of this time, the study has followed children to 7 years of age and found no emotional, behavioural, or cognitive deficits associated with light drinking at any of the time points assessed[9][10][11].  In fact, in the research to date, there has been a J-shaped curve with the light drinkers having children who outperformed the abstainers, even after controlling for the potential confounds (but we’ll talk more about this in a moment).  Further, there were no effects when the pregnancy was planned or not, suggesting that the timing of light alcohol use was not an issue.

The second large-scale, prospective study comes from Australia and followed children up to age 14, an important step as much of the research on FASDs suggests that deficits may not become apparent until adolescence.  In this study, the drinking groups were as follows: abstainers, occasional drinking (up to one standard drink per week), light drinking (2-6 units per week), moderate drinking (7-10 units per week), and heavy drinking (11 or more units per week).  Child outcomes were assessed at birth and then ages 1, 2, 3, 5, 8, 10, and 14 on a variety of child behaviours, both externalizing and internalizing.  In addition, various maternal sociodemographic information (e.g., maternal education, family income, etc.), maternal smoking, and family functioning were included as covariates.  Across all time points, there was no evidence of any internalizing or externalizing problems associated with alcohol use in the first trimester (there was an odd finding of greater externalizing with only occasional drinking later in pregnancy, but given the lack of an effect for heavy drinking, it is likely an issue of type 2 error).  In fact, after controlling for the various confounds, the children in the light drinking category showed fewer behavioural problems across development.

The major issue that arises in these studies is really not just one issue, but a chaotic mix of statistical problems.  Namely, the light drinking group is consistently more advantaged than the abstaining groups and has different drinking patterns.  So abstaining and heavy drinking are more associated with lower SES, intelligence, nutritional status, and so on whereas light drinking is associated with higher levels of these variables.  Now, some have referred to the inability to control for this as simply reflecting “residual confounding”.  This means that although researchers do regularly control for some of these factors (and these longitudinal studies have, as I’ll discuss below), they can’t fully control for everything and variables like nutritional status (which go along with SES) often aren’t included.  What we really have is not only residual confounding, but also disparate subgroups, which means that the relationships between drinking and various outcomes are different for these different groups; in this case, we would expect different relationships by SES, which seems to be what we find.

However, this then implies that the relationship between drinking and various outcomes isn’t the same across these groups.  If it’s biological, wouldn’t we expect the same?  Yes, but also no.  And this brings us to statistical issue number 3 (after residual confounding and disparate subgroups): Third variables.  It is highly likely that the relationship between drinking and outcomes is moderated (or possibly even mediated) by other variables, such as nutritional status, genetics, or something else.  So we have a mix of almost all the major statistical elements that influence relationships and these can all affect the relationship in almost any way so we can’t say that the research is inherently conservative or liberal in the findings.  Now, the longitudinal studies have controlled for many of these confounds, but that’s not enough.

The UK Millenium Study did attempt the best they could do in the analysis at age 7[11] by using a quasi-experimental design which matched individuals on a range of SES factors between the two groups and made comparisons between these groups.  This means that we are looking at the difference in outcomes when people are similar on the confounds instead of just statistically controlling for it, and again, some of the same benefits were found, but many did disappear (though no deficits became apparent).  BUT this also is imperfect as it can’t look at the potential interactions or third variables and still may not fully control for the confounding effects, particularly of genetics, and this was beyond the scope of these longitudinal studies.

Genetic Studies

One of the major findings of recent years is a set of alleles that are associated with the absorption and use of alcohol.  As such, some studies have started looking at drinking outcomes based on the presence or absence of these alleles.  One of the first such study was also the first to show a detriment on IQ at age 8 of light drinking (defined as 1-6 units of alcohol per week) based on the presence of these particular alleles[12], with a greater detriment based on the presence of more alleles known to influence the ability to metabolize alcohol.  Basically, the researchers examined the impact of these alleles on IQ for both abstainers and mothers who reported drinking lightly separately and found a negative effect of alleles for the light drinkers, but not for the abstainers.  This is suggestive but not conclusive.

Needless to say, further research including genetics has been done using what is called a “Mendelian randomization approach”.  This is a method used when there are so many confounds that groups cannot be adequately compared and so a third genotypic variable hypothesized to be related to the variable of interest (e.g., the genotype is related to alcohol consumption), but not related to the various confounds (e.g., the genotype is not related to SES or other potential confounds), is used as the measure instead.  In the work on prenatal alcohol exposure, a specific genotype (ADH1B) is often used because it is associated with alcohol metabolism, is randomized during meiosis, and is unrelated to the various lifestyle factors that are confounds in much of the research.  Using this method, one study examined late prenatal alcohol exposure in predicting asthmas and hay fever and found, contrary to the association based on drinking status, there was no relationship when examined using Mendelian randomization[13].

Another study[14] looked at cognitive development using the Mendelian approach and examined data for those who were moderate drinkers (<7 units of alcohol/week) independent of the group.  Two outcome variables included a standardized IQ test and scores on a national academic test.  With respect to IQ, there were no effects of light drinking using this approach (i.e., children of women without the rare allele did not fare worse than those who carry it), but there was a difference on the academic test.  With the academic test, the children of mothers who carried the rare allele (and thus were more likely to abstain) scored, on average, 1.7 points higher.  However, this result actually included the heavy drinkers who had the largest effect size, though the effect might still hold otherwise, but for this study, even the authors concluded, “[T]here was no robust statistical evidence of genotype x alcohol interaction, assuming a linear trend for drinking categories.”

At this stage we are left with some suggestion that perhaps there is a genotype x alcohol interaction, but we still don’t have the research to say anything conclusively.  Everyone would be wise to keep an eye out for the results of future studies, though even these methods are not bulletproof as the randomization does still include individuals with the rare genotype who still drank, just significantly less so than those with the more common genotype.  What we do have though is evidence that the benefit to light drinking seen in the longitudinal studies is likely due to other confounding factors and should not be used to justify inherent benefits to light drinking.

Meta-Analyses

There have actually been several meta-analyses or systematic reviews on the issue of light drinking.  With respect to pregnancy and birth outcomes, one systematic review[15] considered miscarriage, stillbirth, intrauterine growth restriction, prematurity, birthweight, small for gestational age, and birth defects as outcomes and found no evidence of adverse effects of alcohol use at less than 84 g/week.  This is in line with an earlier meta-analysis looking at fetal malformations which found no effect of drinking up to 2 drinks/day[16].  Another more recent meta-analysis and review also found no effect of prenatal alcohol exposure up to 1 drink/day on low birthweight, preterm birth, and small for gestational age[17].

With respect to cognitive and behavioural outcomes, there are two meta-analyses in recent years.  One meta-analysis looked at infant mental development[18] and found no effects at ages 6-8 months or 18-26 months, but did find weak effects at 12-13 months (even when adjusted for covariates).  Notably, most of the studies were done on low SES families and may reflect interactions with other environmental factors (remember the point above of low-SES families faring worse in the studies examining prenatal alcohol exposure).  Another meta-analysis looking at neuropsychological outcomes[19] (i.e., academic performance, attention, behaviour, cognition, language skills, memory, and visual and motor development) found effects for binge and moderate drinking, but no effect of mild prenatal alcohol exposure.

Of course, all of these meta-analyses are only as good as the research inside and the research often suffers the same statistical problems discussed above and the methodological issues below.

Methodological Problems

Defining Light Drinking

One of the biggest problems comes in the definition of “light drinking”.  As you may have already gleaned from the readings above, many studies include up to 1 unit of alcohol per day as “light” whereas other research considers it to be 2-3 units of alcohol per week.  With differing definitions of “light drinking”, it becomes difficult to make comparisons or even to know what we’re talking about when talking about light drinking.

The second issue is the understanding of a unit of alcohol.  In research is it pretty clear cut in that one unit is a specific amount, but individuals taking part in research and who are pregnant may not realize what this amount is.  Take the Queensland study, for example, where the conclusion is “2 glasses of wine” yet the measured amount is 4 units, or glasses.  Why the discrepancy?  Because in the UK, the average wine served in a restaurant is closer to 2 units, or 300ml (1 unit is 150ml of wine).  Thus the number of units per glass will depend on where you are, but 4 units is just shy of 1 bottle of wine, so whether this is defined as “heavy” or “moderate” drinking remains up for debate.  It certainly isn’t “light” drinking, though.  However, as research on alcohol use depends upon self-report, when people don’t understand how much alcohol is in a unit versus a glass they might be served at a bar, they can often underreport their alcohol use.  (Of course, for the purposes herein, underreporting actually serves as a liberal bias which would mean we’d be more likely to see effects of “light” drinking.)

 

Genetics

Some research, as discussed above, does seem to suggest that certain genes may influence the likelihood that someone who drinks at all will have a child with a deficit (though even that is mixed).  It is also worth noting that according to one research review, only 6% of children of alcoholics will develop FAS.  This suggests that there are other factors – likely some heritable – that influence the likelihood of being negatively impacted by prenatal alcohol exposure.  Research into these factors is a must as anything that helps identify higher-risk individuals is a good thing.

Confounding

There are a few confounding variables that should be considered and yet aren’t always.  In US research, for example, we know that women with higher IQ and education are more likely to drink which may mask the effects of “light” drinking on certain outcomes.  Some research does control for this, but not all does.  This is particularly important for any outcome that is expected to be related to SES or maternal IQ – such as child cognitive development.  This should not, however, matter for physical markers of FAS or FASD and theoretically, it’s hard to suggest it should matter for certain behavioural outcomes like hyperactivity.

In addition, certain variables that matter to all non-physical outcomes are often ignored, one notable example is maternal (and paternal) mental health.  When included, mental health has found to be far more predictive of behavioural outcomes often attributed to prenatal alcohol exposure yet it may be a third variable situation in which mental health affects both drinking and the child outcomes.  Another example is prenatal (and postnatal) nutrition[20]. This difference in nutrition may also explain why there are clear SES and race relationships with FAS with lower rates in the offspring of middle class mothers as opposed to lower class mothers, despite higher drinking rates reported for middle class women.

There are many variables to consider as drinking alcohol (especially larger amounts) rarely occurs in the absence of other risk factors. These factors will be more or less relevant depending on the outcome of interest and may serve as independent risk factors and/or ones that interact with alcohol exposure to increase the risk of negative effects.

Alcohol Timing

When people drink matters a lot, as much as the amount.  First trimester drinking is highlighted as the worst time in terms of outcomes, which causes concern amongst women who drank before knowing they were pregnant and is what led to the patronizing CDC recommendations.  What we do have, though, does suggest that drinking in the first half of the first trimester is not high-risk and in fact is more “all or none” with enough damage to cause a miscarriage or nothing at all.  This makes sense theoretically given that implantation hasn’t occurred and thus transfer of nutrients from mother to fetus is all but absent (it occurred earlier).  This is one of the problems with animal models as their day 1 of gestation is the equivalent of approximately 4 weeks post-fertilization for humans.  Unless we study human alcohol timing, we can’t glean a lot of information from animal research.

Mechanisms

One of the things that has to occur with research is a plausible mechanism.  Researchers are starting to identify the epigenetic changes that take place with exposure to moderate alcohol using animal models which gives us ideas as to how FAS can occur.  With FASD, we are missing the “how” for many of the assumed links and this may be where certain third variables, like maternal mental health, come into play, but more research is needed.

Conclusions

I leave it up to everyone to read the findings and make their own decision on the risk of light drinking (and taking the most conservative estimate, it means 1-2 drinks per week), knowing that, yes, the only way to absolutely avoid FASD is to not drink at all.  Discussing the research does not mean that anyone is suggesting people should drink and especially does not speak to drinking more than lightly.

But what if you choose to drink?

The evidence does not support the demonization of women who choose to have the occasional drink and luckily associations like NOFAS also don’t support any legislation that would charge women for making this choice.  The lack of known risks means that we cannot judge a woman for this occasional drink as the research continues to fall in one direction, though clearly isn’t wholly complete yet.  Education is key and with education, women can make their own decisions.  Notably, education doesn’t just come in the form of fear-mongering, but in providing the evidence as well as any needed assistance.

When it comes to public policy, I understand that the aim is for the greatest reduction in harm, but I personally don’t see the push to demonize light drinking as actually helping on a large-scale compared to other possible interventions.  Light drinkers are not at risk so far as we can tell right now (future research notwithstanding), despite research really wanting to find something wrong, and so efforts to reduce their drinking may yield no benefits.  Moderate to heavy drinking is, and when we focus efforts on just saying “don’t drink” we miss the people most at risk.  Why?  Because there are almost always underlying reasons for the drinking they do and we are failing dramatically in other areas – such as mental health and nutrition associated with poverty – in our society.  I would much rather see a brief intervention of mental health screening at all doctor’s visits (regardless of pregnancy) with proper referrals for those who score above a given cut-off than checking in on drinking behaviour (which is likely lied about anyway).  This would serve to inherently reduce alcohol use all around, but also may help those children whose behaviours are being driven by mental health problems in the home, regardless of alcohol use (which research does tell us is a far bigger problem than light drinking).

One thing is for sure: We need more research into the risk and protective factors that exist if we are to help as many families as possible.

_________________________________

[1] Alati R, Smith GD, Lewis SJ, Sayal K, Draper ES, et al. Effect of prenatal alcohol exposure on childhood academic outcomes: contrasting maternal and paternal associations in the ALSPAC study.  PLoS One 2013; doi: 10.1371/journal.pone.0074844.

[2] Feldman HS, Jones KL, Lindsay S, Slymen D, Klonoff-Cohen H, et al. Prenatal alcohol exposure patterns and alcohol-related birth defects and growth deficiencies: a prospective study.  Alcoholism: Clinical and Experimental Research 2012; 36: 670-6.

[3] Andersen AMN, Andersen PK, Olsen J, Grønbæk M, Strandberg-Larsen K.  Moderate alcohol intake during pregnancy and risk of fetal death.  International Journal of Epidemiology 2012; doi: 10.1093/ije/dyr189.

[4] Mills JL, Graubard BI.  Is moderate drinking during pregnancy associated with an increased risk of malformations?  Pediatrics 1987; 80: 309-14.

[5] Bailey BA, Sokol RJ.  Prenatal alcohol exposure and miscarriage, stillbirth, preterm delivery, and sudden infant death syndrome.  Alcohol Research & Health 2011; 34: 86-91.

[6] Larkby C, Day N.  The effects of prenatal alcohol exposure.  Alcohol, Health, & Research World 1997; 21: 192-8.

[7] Day NL, Richardson GA, Geva D, Robles N. Alcohol, marijuana, and tobacco: The effects of prenatal exposure on offspring growth and morphology at age six. Alcoholism: Clinical and Experimental Research 1994; 18:786-794.

[8] Fried PA, O’Connell CM. A comparison of the effects of prenatal exposure to tobacco, alcohol, cannabis and caffeine on birth size and subsequent growth. Neurotoxicology and Teratology 1987; 9:79-85.

[9] Kelly Y, Sacker A, Gray R, Kelly J, Wolke D, Quigley MA.  Light drinking in pregnancy, a risk for behavioural problems and cognitive deficits at 3 years of age?  International Journal of Epidemiology 2009; 38: 129-40.

[10] Kelly YJ, Sacker A, Gray R, Kelly J, Wolke D, et al. Light drinking during pregnancy: still no increased risk for socioemotional difficulties or cognitive deficits at 5 years of age?  J Epidemiol Community Health 2010; doi: 10.1136/jech.2009.103002.

[11] Kelly Y, Iacovou M, Quigley MA, Gray R, Wolke D, et al. Light drinking versus abstinence in pregnancy – behavioural and cognitive outcomes in 7-year-old children: a longitudinal cohort study.  BJOG 2013; 120: 1340-7.

[12] Lewis SJ, Zuccolo L, Smith GD, Mcleod J, Rodriguez S, et al. Fetal alcohol exposure and IQ at age 8: evidence from a population-based birth-cohort study.  PLoS One 2012; 7: e49407.

[13] Shaheen SO, Rutterford C, Zuccolo L, Ring SM, Smith GD, et al.  Prenatal alcohol exposure and childhood atopic disease: a Mendelian randomization approach.  Journal of Allergy and Clinical Immunology 2014; 133: 225-32.

[14] Zuccolo L, Lewis SJ, Smith GD, Sayal K, Draper ES, et al.  Prenatal alcohol exposure and offspring cognition and school performance. A ‘Mendelian randomization’ natural experiment.  International Journal of Epidemiology 2013; 42: 1358-70.

[15] Henderson J, Gray R, Brocklehurst P.  Systematic review of effects of low-moderate prenatal alcohol exposure on pregnancy outcome.  BJOG 2007; 114: 243-52.

[16] Polygenis D, Wharton S, Malmberg C, Sherman N, Kennedy D, et al.  Moderate alcohol consumption during pregnancy and the incidence of fetal malformations: a meta-analysis.  Neurotoxicology and Teratology 1998; 20: 61-7.

[17] Patra J, Bakker R, Irving H, Jaddoe VWV, Malini S, Rehm J.  Dose-response relationship between alcohol consumption before and during pregnancy and the risks of low birthweight, preterm birth and small for gestational age (SGA)—a systematic review and meta-analysis.  BJOG 2011; 118: 1411-21.

[18] Testa M, Quigley BM, Eiden RD.  The effects of prenatal alcohol exposure on infant mental development: a meta-analytical review.  Alcohol and Alcoholism 2003; 38: 295-304.

[19] Flak AL, Su S, Bertrand J, Denny CH, Kesmodel US, Cogswell ME.  The association of mild, moderate, and binge prenatal alcohol exposure and child neuropsychological outcomes: a meta-analysis.  Alcoholism: Clinical and Experimental Research 2014; 38: 214-226.

[20] Young JK, Giesbrecht HE, Eskin MN, Aliani M, Suh M.  Nutrition implications for fetal alcohol spectrum disorder.  Advances in Nutrition 2014; 5: 675-92.