Noble Lies: alcohol and pregnancy edition

Light drinking during pregnancy - say a glass of wine every second or third day - really doesn't seem to do any harm. One pretty thorough study in 2010 showed that kids of mothers who drank moderately during pregnancy had better outcomes than those born to abstainers. I expect that most of that effect is due to unobserved cohort heterogeneity, but it would be pretty tough for that confound to reverse the sign of the effect. I'd summarized:

First, some summary data. About eight percent of their sample - just over 900 mothers - reported moderate or heavy drinking during pregnancy. Incidence was slightly skewed towards lower education cohorts, with just over three percent of low education mothers reporting heavy or binge drinking as compared to just over two percent of mothers in the top educational category. Highly educated mothers were the ones most likely to report light drinking during pregnancy, though the modal highly educated mother reported absention during pregnancy. Among the lowest educational cohorts, absention during pregnancy was by far the most likely reported behaviour. So if you're a highly educated mother who sneaks a drink occasionally, you're certainly not alone.

They ran regressions on child outcomes controlling for maternal drinking and for the confounds. The unadjusted data showed massive benefits to being born to a light-drinking mother, but that was mostly due to the confounds; results attenuated in the full models, as you'd expect if it's the higher education parents who are more likely to have a little but not a lot. Controlling for the parental factors didn't reverse the effects, but it really knocked back the magnitude and statistical significance. Typical coefficients on cognitive benefits to the child of the mother's light drinking were knocked back to a third of their baseline magnitude with the full set of controls; it's consequently more than plausible that remaining uncontrolled things that differ between "not during pregnancy" mothers and light-drinking mothers could be responsible for at least some of the remaining gap. An IV strategy would be needed to estimate things more precisely.

But the bottom line seems to be no bad effect, and some (small) chance of a positive effect, of a pregnant woman's drinking up to a drink or two "per week or per occasion" instead of abstaining during pregnancy. I wouldn't bank on the positive effects holding up to an IV model, but I'd also be very surprised if those techniques drove things to a negative effect.

The usual bunch gave a lot of warnings about "no safe level", along with some reasonable critiques of some of the sillier journalistic takes on the study - like those that reported uncorrected differences in means rather than coefficient estimates. My best takeaway from the paper is that low level drinking is very unlikely to do harm, but heavy drinking is very likely to do harm.

Another set of papers is out showing much the same thing: low levels of drinking, or even occasional binge drinking, doesn't seem to do much harm, but heavy drinking does. And the usual bunch gave the usual party line:

Christine Rogan, Health Promotion Advisor (Coordinator of the Fetal Alcohol Network NZ), Alcohol Healthwatch, comments:

“These types of studies do not provide a reason for pregnant mums to pop the champagne! Sadly though that is often the effect they have. The public (media) love good news headlines about drinking and the headline always wins, no matter the fine print such as the authors themselves conclude that no safe level of consumption has been established.

I'm not sure what gets to constitute proof that something is safe if repeated findings of "no more risky than abstinence" don't count. And Rogan is wrong when she says that results came only from teachers' reports on the students rather than from testing of the students themselves: this paper in the set, at least, included measures of the child's IQ at age 5.

Otago's required evidentiary standards seem to depend a lot on the conclusion. If you want to say anything about whether women should be a bit more relaxed about having a drink every two or three days during pregnancy, nothing outside an ethically forbidden randomized control trial seems to suffice. If you want to ban smoking outside of bars, surveying 13 people you met on Facebook is enough.

I can't help but run anything coming out of Otago's healthists through the "noble lie" filter: are they saying what they really believe, or are they scared that idiots will take findings of "light drinking does no measurable harm" as justification for very heavy and harmful drinking during pregnancy?

Dave Guerin at Ed.co.nz seems to agree; he summarizes:

Alcohol in Pregnancy Danish research on the impact of alcohol during pregnancy was interpreted in quite heroic ways by academics to support the current public health view in NZ (SMC, NZ Herald, Fairfax, One News).