From the perspective of one who has repeatedly urged caution in overstating the dangers of drinking during pregnancy [1], the present paper [2] presents a reasoned attempt to add clarity to key issues which, unfortunately, do not lend themselves to any clear-cut resolution. As my space is limited, I will focus on what I consider to be the most refractory of these: the intertwined issues of threshold and definition of ‘low-to-moderate’ drinking.

First, in reverse order. The authors frequently allude to ‘low-to-moderate’ drinking but never define what they mean. Because they do not, their comments are not as compelling as they might otherwise be. For example, they state ‘evidence from animal experiments’ to the effect that ‘low-to-moderate levels of alcohol exposure in pregnancy may be associated with effects on the developing brain’, citing Ikonomidou et al.'s [3] study of apoptotic degeneration in rats. The authors of that study, however, found that no discernable effects occurred in 7-day-old rats if the dose did not produce a blood alcohol level (BAL) of less than 200 mg %, a BAL far above the legal limit for being charged with driving while intoxicated in every inch of the United States. In human terms, it is the equivalent of 11 drinks consumed over a 5-hour period: hardly low-to-moderate alcohol consumption. As Gray and his colleagues note, the evidence of adverse effects, however classified, is not very clear for drinking at low-to-moderate levels and for binge drinking in women who do not otherwise drink heavily. Had they realized how high the BALs were in the Ikonoimodou study, they may have been even more forceful in their conclusion.

Secondly, how is a threshold determined? In most reports involving human consumption of alcohol, amounts are reported typically in terms of average drinks per week or month; but like the proverbial 2.3 children per family, the statistical average drinks says little about what is actually consumed per drinking occasion. Nor does it say by whom (about which more in a moment). More often than not, an average of seven drinks per week is seven drinks on a Friday or Saturday night. Averages are summaries—they convey some information, but with respect to thresholds this is misleading information, because it is the BAL that is most relevant to whatever outcome is being measured. Surely, in the absence of BAL, the critical element in deciding what is a safe level of drinking is the actual number of drinks per drinking occasion that is consumed during a pregnancy that matters. That is something that is and, I suspect, will continue to be, very difficult to translate into a ‘threshold’ below which drinking is ‘safe’, and above which it is not.

Animal studies are useless for establishing ‘safe’ thresholds; in fact, anything of relevance to the human situation other than alcohol in huge amounts, usually above 150 mg % [4,5], can produce statistically significant effects. A rat intubated with a dose of 6 g of alcohol per kilogram body weight per day is being given the human equivalent of 27 drinks per day. An animal put on a liquid alcohol diet consumes about 12 g per kg per day, which is the human equivalent of 54 drinks per day [1]. Moreover, these are not one-time exposures but are maintained for several days, the equivalent of a chronic binge lasting almost an entire pregnancy.

Determining and putting into some kind of usable classification of drinks per occasion for statistical analysis of threshold determinations is only another part of the proverbial tip of the iceberg conundrum (as an aside, there appear to be so many tips of the iceberg associated with fetal alcohol syndrome that the tip must surely have grown to a full-fledged iceberg by now). Numerous studies have noted that, even though they may supposedly drink the same, some women give birth to children with ‘fetal alcohol spectrum disorder’ and some do not. A reductionist approach focused on alcohol alone is misguided [6], because fetal alcohol syndrome or spectrum or whatever the fashionable term is, or may be tomorrow, is not an equal opportunity disorder—it affects mainly destitute women and is nearly always found among women who smoke (also associated with poverty) [7]. The extremely high rates of fetal alcohol syndrome that the authors mention in studies from South Africa, Italy [8,9] and elsewhere come from the poorest segments of those countries. Alcohol is certainly involved, but what does that tell us about consumption of one or even two drinks at dinner in more affluent countries?

Let us concede that heavy drinking (five or more drinks per occasion) over an extended period of time is not good for anybody—mother, father or fetus. Let us also concede that the research claiming ‘adverse effects’ of low-to-moderate drinking is, as Gray and his colleagues acknowledge, not convincing.

Declaration of interest