Critique 137: Alcohol intake prior to and during pregnancy and birth outcomes — 28 March 2014
Nykjaer C, Alwan NA, Greenwood DC, Simpson NAB, Hay AWM, White KLM, Cade JE. Maternal alcohol intake prior to and during pregnancy and risk of adverse birth outcomes: evidence from a British cohort. J Epidemiol Community Health 2014, pre-publication. doi:10.1136/jech-2013-202934
Authors’ Abstract
Background: Evidence is conflicting regarding the relationship between low maternal alcohol consumption and birth outcomes. This paper aimed to investigate the association between alcohol intake before and during pregnancy with birth weight and gestational age and to examine the effect of timing of exposure.
Methods: A prospective cohort in Leeds, UK, of 1303 pregnant women aged 18–45 years. Questionnaires assessed alcohol consumption before pregnancy and for the three trimesters separately. Categories of alcohol consumption were divided into ≤2 units/week and >2 units/week with a nondrinking category as referent. This was related to size at birth and preterm delivery, adjusting for confounders including salivary cotinine as a biomarker of smoking status.
Results: Nearly two-thirds of women before pregnancy and over half in the first trimester reported alcohol intakes above the Department of Health (UK) guidelines of ≤2 units/week. Associations with birth outcomes were strongest for intakes >2 units/week before pregnancy and in trimesters 1 and 2 compared to non-drinkers. Even women adhering to the guidelines in the first trimester were at significantly higher risk of having babies with lower birth weight, lower birth centile and preterm birth compared to non-drinkers, after adjusting for confounders (p<0.05).
Conclusions: We found the first trimester to be the period most sensitive to the effect of alcohol on the developing fetus. Women adhering to guidelines in this period were still at increased risk of adverse birth outcomes. Our findings suggest that women should be advised to abstain from alcohol when planning to conceive and throughout pregnancy.
Forum Comments
There have been literally hundreds of epidemiologic studies relating the maternal use of alcohol during pregnancy and outcomes among the offspring. There is no question that heavy alcohol use, generally among women who are known alcoholics, can be associated with a very serious group of abnormalities known as Fetal Alcohol Syndrome (FAS). Further, some, but not all, follow-up studies of pre-natal alcohol use by women have shown some physical and/or behavioral defects that have been attributed to lesser degrees of maternal alcohol use, and classified as Fetal Alcohol Exposure (FAE). A persisting difficulty is that a known percentage of neonates will demonstrate some abnormality; these are often classified as FAE only if there is evidence of alcohol use by the mother during pregnancy, but otherwise blamed on some other exposure.
According to some, no level of pre-natal alcohol consumption has been shown to be entirely free of adverse effects on the fetus, so the common recommendation in many countries is for no alcohol consumption at all during pregnancy. Some scientist s have argued that this is unnecessarily conservative, as many developmental studies of infants and children have not shown differences according to whether or not the mother consumed light or occasional alcohol during her pregnancy.
The present study, while not large (1,303 pregnant women) evaluated reported alcohol intake to birth weight and the risk that an infant is small for gestational age (SGA). The authors concluded that, in this group of English women, there was an increased risk of an infant having a low birth weight and being SGA if the mother reported any alcohol intake during pregnancy, especially during the first trimester. The study also related the reported alcohol intake in the 4 weeks prior to the woman became pregnant, and found an increased risk for women who consumed alcohol and the risk of the infant being SGA.
Previous research on maternal alcohol exposure during pregnancy: There is a serious problem in interpreting all studies of alcohol and the outcomes of pregnancy because of potential confounding, especially by level of education, ethnicity, smoking, coexisting use of illegal drugs, etc. Further, alcohol use by pregnant women has become an especially emotional and even a moral issue. As stated by reviewer Finkel, “This subject is an emotional and scientific morass, and something of a minefield. In addition, the entangled confounders make rational analysis most difficult. Here, indeed, is a challenge for the epidemiologists and statisticians. The literature can support varying views, from no alcohol at all for those who might soon be expectant mothers and fathers to drinking moderately throughout pregnancy.”
Forum member Stockley had similar concerns: “The relationship between alcohol consumption and pregnancy outcomes other than FAS appears controversial and uncertain, so it is difficult to draw any conclusion and give absolute advice, although a maximum daily amount and pattern to minimize risk should be advised.
“For every paper that claims that any alcohol negatively influences a birth outcome, another paper refutes it: where Kesmodel et al (Epidemiology 2000;11:512-518) observed an increased risk of preterm delivery only with consumption of larger amounts of alcohol, Albertsen et al (Am J Epidemiol 2004;159:155-161) did not with four drinks per week over the gestation period. An increased risk of preterm birth was also not observed by Parazzini et al (European J Clin Nutr 2003;57:1345–1349) until more than three drinks on average per day were consumed.
“Concerning low birth rate, while Covington et al (Neurotoxicology Teratology 2002;24:489-496) observed that more than 14 alcoholic drinks/week decreased birth weight and length, and was associated with lower weight at age seven years, O’Callaghan et al. (Early Hum Dev 2003;71:137-148) did not observe this. Furthermore, Mariscal et al (Ann Epidemiol 2006;16:432-438) observed that alcohol consumption of less than 6 g/day actually decreased the risk for low birth weight, but the risk was increased when more than 12 g/day of alcohol was consumed. The risk was decreased again when the 12 g/day was confined to weekends for non-cigarette smoking women. This also shows the confounding of cigarette smoking and the importance of the pattern of alcohol consumption.
“Concerning spontaneous abortion, while Kesmodel et al (Alcohol and Alcoholism 2002;37:87-92) observed an increased risk of spontaneous abortion when five or more alcoholic drinks/week were consumed in the first trimester, which was corroborated by Henriksen et al (Am J Epidemiol 2004;160:661–667) but at 10 or more alcoholic drinks/week, but this was not found by Maconochie et al (2006).
“Concerning neurobehaviour and cognition, O’Callaghan et al (Early Hum Dev 2007;83:115-123) did not observe any adverse attention, learning or cognition outcomes when less than one alcoholic drink/day was consumed but drinking more than this in late pregnancy and indeed binge drinking, was associated with an increased risk of overall learning difficulties. D’Onofrio et al (Arch Gen Psychiatry 2007;64:1296-1304) stated that polydrug use during pregnancy was a better indicator of behavioural and learning difficulties.” As stated by Forum member Skovenborg: “The issue of maternal alcohol intake prior to and during pregnancy has sunk into the quick sands of political science, poor science, serious confounding and large problems with bias including information bias (self report on alcohol intake) and that situation is not going to change for the better in the near future.”
Contributions to the topic by the present paper: Forum members have detected a number of potential problems with the present paper. The number of subjects is quite small for evaluating such a relationship. Further, the majority of eligible subjects did not agree to take part in the study (only 30% agreed), so there may be limited application of the results to the general population. Also, these women in Leeds, at least those who agreed to take part in the study, were much more likely to be drinkers than in most previous studies of pregnant women. In the 4 weeks prior to pregnancy, 74% of women reported ≥ 2 units/week, with a mean intake of 15.1 units per week. During the first trimester, 78% of the women reported alcohol consumption, and 53% of the women reported consuming ≥ 2 drinks/week. It is also interesting that for first trimester alcohol consumption, the effects on birth weight, being SGA, and having a preterm birth did not show a dose-response curve; the women reporting ≤ 2 drinks/week had essentially the same effects as those who reported that they consumed more.
It is unfortunate that the investigators did not include the pattern of drinking (regular versus binge drinking) in their analyses. During the first trimester, the average intake of alcohol in the ≥ 2 drinks/week group was 7.2 drinks; consuming this number of drinks on a single occasion during the week would have very different effects on blood alcohol levels (and presumably on the fetus) than if one drink was consumed daily. The authors state that the women in this study had a variety of ethnic backgrounds, but do not discuss potential differences in effect of alcohol when data were stratified by ethnicity. In the US, there are very large differences in birth weight and an infant being SGA among most minority populations, when compared with women of European ethnicity, and residual confounding could be expected if only a variable for ethnicity was included in the multi-variable analysis, as was done in this study.
Reviewer Zhang had some more specific problems with the study. “Overall the results are difficult to comprehend. For example, the crude difference in birth weight between non-drinkers and drinkers assessed 4 wks prior to pregnancy was relatively small (-14.6 g with <2 unit/wk, and -23.2g/wk with >2 units/wk); however, after adjusting for potential confounders, the differences were substantially increased (-70.2g/wk and -105g/wk, respectively), indicating that some strong confounders existed between alcohol consumption and birth weight. These confounders should have a strong protective effect on birth weight and were more common among alcohol drinkers. In this study, compared with non-alcohol drinkers, women who drank >2 units/wk were older, smoked less, had a higher percentage of college education, were more likely of European origin, and had a higher percentage of being primigravida. On the other hand alcohol drinkers consumed substantially more coffee (a risk for low birth weight) than non-consumers of alcohol. The data only present the mean and 95% CI for coffee drinking. If coffee consumption between alcohol drinkers and non-drinkers did not overlap, then simply adding coffee drinking into the regression model will not properly adjust for the confounding effect of coffee drinking on the association between alcohol consumption and low birth weight (violation of positivity assumption). Similar issues can be applied for BMI at pre-pregnancy.
“Further, the average of alcohol consumption 4 weeks before pregnancy was 15.1 units/wk for those who consumed >2 units/wk (95% CI 14.1-16.1), suggesting that more than 97.5 % of women in that group consumed >2 units/day of alcohol. However, among those who consumed <2 units/wk, the 95% CI of alcohol consumption was 0.9-1.1 units/wk, suggesting very few women consumed alcohol between 2 units/wk to 14 units/wk. Such a distribution of alcohol consumption found in this group is a little bit odd. This finding only can be interpreted if women who drank more than 2 units/day had an increased risk of having a low birth weight infant to begin with, instead of it being based on their consumption of >2 units/wk.
Finally, epidemiologic studies of pregnancy outcome can only study prevalent outcomes. If risk of spontaneous abortion is unevenly distributed between comparison groups, it would bias the effect estimates. For example, if non-drinkers had higher risk of spontaneous abortion due to their other risk factors, then it would bias the effect estimate of alcohol consumption and low birth weight away from the null. Further, there is not mention of C-section in the paper, which also relates to birth weight. My comments neither endorse nor disapprove the conclusions of the current study findings, but rather raise several methodological issues and raise question as to how appropriately to interpret the reported findings.”
Forum member Thelle focused his comments on three topics: selection bias, confounding, and the public heath message. He states: “If selection bias should be the explanation for the observed effect, then the association between alcohol and birth weight among the non-attendees should be the opposite, assuming that there is no true association between alcohol and birth weight. That the selective forces should be acting in this direction is unlikely. As for confounding, the consumption of coffee has been raised. This is of interest as coffee during the first trimester is associated with increased risk of fetal deaths, but nausea is at the same time a protective factor resulting in reduced coffee intake. Finally, the public health concern differs in this situation from that of health effects of alcohol on the adult population. The fetus is unprotected and has no will in determining whether to be exposed to alcohol or not. This is neither dogmatic nor emotional, and the message to the public must be based on the consequences of giving the wrong advice. It is in my view better to err on the safe side in this particular situation.”
Forum Summary
The present analysis was carried out among 1,264 women from Leeds, UK, whose alcohol intake was estimated prior to and during pregnancy. The outcomes were birth weight of the infant and whether or not it was small for gestational age (SGA). In comparison with most previous research on this topic, this study is notable for being of a relatively small size. Further, it had a high percentage of women who consumed alcohol both prior to and during the first trimester, had a very low participation rate of eligible women (only 30%), failed to present estimated effects by ethnicity, and had some questions related to residual confounding. It is unclear how the results of this study increase our knowledge of maternal alcohol consumption and birth outcomes.
Forum members do not believe that the consumption of alcohol should be recommended for pregnant women. A certain percentage of newborns will be small for gestational age, have certain deformities, and have later emotional and behavioral abnormalities. If a woman has consumed any alcohol during the pregnancy, she (and perhaps even her doctor) may blame the abnormality on alcohol consumption, whether or not it had anything to do with it. Hence, the majority of Forum members agree that women should not be encouraged to consume alcohol during pregnancy.
Further, scientific data are very consistent on the potential risks of serious adverse health outcomes of the infant from heavy maternal drinking, especially among women who are alcoholics. Heavy-drinking women who become pregnant should be strongly urged to stop their drinking.
On the other hand, sound data indicating harmful effects on the fetus of light or occasional drinking by a pregnant woman are difficult to come by. Because of the epidemiologic concerns of the present paper, described in our Forum critique, we do not believe that this study adds materially to our understanding of the topic. The serious anxiety occurring among some women who may have ingested some alcohol prior to learning that they were pregnant seems, based on numerous studies, to generally be unnecessary; the need to abort a fetus because of previous light drinking by the mother cannot be justified.
There is a serious problem in interpreting all studies of prenatal alcohol exposure and the outcomes of pregnancy because of potential confounding, especially by ethnicity, education, smoking, coexisting use of illegal drugs, etc. Further, alcohol use by pregnant women has become an especially emotional and even a moral issue, and the entangled confounders make rational analysis most difficult. This means that there is a difficult challenge for epidemiologists and statisticians, as well as for public health officials and the general public, about how to interpret results of individual research projects. The literature can support varying views, from no alcohol at all for those who might soon be expectant mothers to being able to drink moderately throughout pregnancy.
Also, it should be pointed out that the potential health benefits of moderate alcohol consumption relate primarily to middle-aged and older people, so there is no reason for pregnant women to consume alcohol for its “health effects.” Thus, it is very reasonable that the majority of women choose to avoid alcohol during pregnancy. Further, heavy drinking during pregnancy has known potentially serious consequences, and should never be encouraged. Finally, there is insufficient scientific evidence that an occasional drink of alcohol during pregnancy leads to harm to the fetus, and should not cause undue alarm in a pregnant woman who may have consumed some alcohol before she realized she was pregnant.
* * *
Comments on this critique by the International Scientific Forum on Alcohol Research were provided by the following members:
Dee Blackhurst, PhD, Lipid Laboratory, University of Cape Town Health Sciences Faculty, Cape Town, South Africa
Giovanni de Gaetano, MD, PhD, Department of Epidemiology and Prevention, IRCCS Istituto Neurologico Mediterraneo NEUROMED, Pozzilli, Italy
R. Curtis Ellison, MD, Section of Preventive Medicine & Epidemiology, Boston University School of Medicine, Boston, MA, USA
Harvey Finkel, MD, Hematology/Oncology, Boston University Medical Center, Boston, MA, USA
Ulrich Keil, MD, PhD, Institute of Epidemiology and Social Medicine, University of Münster, Münster, Germany
Jean-Marc Orgogozo, MD, Professor of Neurology and Head of the Neurology Divisions, the University Hospital of Bordeaux, Pessac, France
Erik Skovenborg, MD, Scandinavian Medical Alcohol Board, Practitioner, Aarhus, Denmark
Creina Stockley, PhD, MBA, Clinical Pharmacology, Health and Regulatory Information Manager, Australian Wine Research Institute, Glen Osmond, South Australia, Australia
Arne Svilaas, MD, PhD, general practice and lipidology, Oslo University Hospital, Oslo, Norway
Pierre-Louis Teissedre, PhD, Faculty of Oenology – ISVV, University Victor Segalen Bordeaux 2, Bordeaux, France
Dag S. Thelle, MD, PhD, Senior Professor of Cardiovascular Epidemiology and Prevention, University of Gothenburg, Sweden; Senior Professor of Quantitative Medicine at the University of Oslo, Norway
David Van Velden, MD, Dept. of Pathology, Stellenbosch University, Stellenbosch, South Africa
Andrew L. Waterhouse, PhD, Marvin Sands Professor, Department of Viticulture and Enology, University of California, Davis; Davis, CA, USA.
Yuqing Zhang, MD, DSc, Epidemiology, Boston University School of Medicine, Boston, MA, USA