1. Doctor, Department of Epidemiology, UCLA—School of Public Health, Katrine Strandberg-Larsen, The National Institute of Public Health, University of Southern Denmark, Denmark. E-mail:
    Search for more papers by this author

Alcohol is neurotoxic in high doses, as well as to the fetal brain, and it is well established that a high intake of alcohol during pregnancy increases the risk of neurological disorders in the child. However, it is still unclear whether it is alcohol or its primary metabolite acetaldehyde that causes harm [1–3]. The existing evidence suggests no detectable harm associated with low intake of alcohol equivalent to less than one drink per day [4]. There may well be a threshold for effect, and it is likely that the fetal brain susceptibility varies largely by gestational age. A recent study indicated, for example, that binge drinking in gestational weeks 11–16 correlated with an increased risk of epilepsy in early childhood [5]. However, as Gray et al.[6] suggest, the introduction of a new diagnostic entity, alcohol-related neurodevelopmental disorder (ARND), could confuse the issue rather than clarifying existing thoughts. It is always a risky matter to include a potential cause/exposure (such as alcohol) into a definition of a disease. By so doing the exposure becomes a necessary cause in the strong sense of the word simply by means of using a circular argument.

Why is it so difficult to reach a consensus? The problem is that life-style factors are correlated closely and are influenced by social, cultural and religious norms.

Many of the studies on high alcohol intake are conducted in populations where such an intake is outside acceptable social norms. These studies focus on women who are socially marginalized and may have many other risk factors, especially poor dietary habits. Some of the published ‘effects’ may well be related, therefore, to an insufficient diet rather than alcohol itself. Studies from some European countries, where alcohol intake in moderate doses during pregnancy is more acceptable than in the United States, may provide better evidence. Studies from Denmark show, for example, that women with a low to moderate alcohol intake during pregnancy tend to come from more socially advantaged groups than abstainers [7,8].

The second concern is the lack of valid exposure data. Being short of good biomarkers of alcohol intake that work in epidemiological studies presents a severe risk of information bias. Again, the better data will come from countries that accept some alcohol intake during pregnancies, but even in this situation alcohol intake is expected to be under-reported, as we have no reason to believe that pregnant women would systematically over-report their consumption of alcohol. Much trust has been given to a possible equal under-reporting of all levels of intake, but that is not realistic. Women who drink only one glass of wine with their dinner and have done so for decades will be able to recall such an intake accurately. Higher, but variable, doses and binge drinking will more often be underestimated. Furthermore, some women do not drink for health reasons and they are subject to confounding by indication. We agree that given these obstacles ecological studies become of interest, especially studies where sudden changes in alcohol intake are seen over shorter time-periods. Ecological studies are placed low in most quality rankings, but such rankings may be grossly misleading in specific situations. However, the critical issue is whether total alcohol consumption correlates with alcohol consumption in pregnant women.

Mendelian randomization is an interesting research option using genes that regulate alcohol metabolism as an instrumental variable for alcohol intake. This instrumental variable is not expected to be confounded because of the random allocation at the time of conception following Mendel's first and second laws. The caveat is canalization; genes regulating metabolism will affect intake because the rate of metabolism is known, to some extent, to the drinker from experience. People who have genes that encode slow metabolism of alcohol tend to drink more than people with genes that encode fast metabolism of alcohol [9]. Thus, it is necessary to stratify on alcohol use and we have to use the often poor data we can obtain by relying upon self-report. A Mendelian randomization study has not yet been conducted, to our knowledge, but sufficient sample sizes are now available and the metabolism genes are well known. We would have more trust in such a study than in looking at time trends of reproductive hazards related to ecological data on alcohol intake that may be too inaccurate for pregnant women.

Declaration of interest