Why We Need to Ban the Word “Safe” from Reports of Risk Research: The Example of Alcohol Use in Pregnancy

by Brian Zikmund-Fisher on June 21, 2012

Yesterday, BJOG: An International Journal of Obstetrics & Gynaecology electronically published in “Early View” 5 articles from a Danish research group, all of which use longitudinal data from 2003-2008 to consider the relationship (or lack thereof) between maternal alcohol consumption during pregnancy and various outcomes (executive function, intelligence, attention) in these mothers’ children at 5 years of age.

The detailed abstracts of the 5 papers are available here, here, here, here, and here. However, I think that one can summarize the gist findings of these studies, when taken as a set, as follows:

  1. Confirming previous research, children of mothers who drank heavily and regularly (>9 drinks per week consistently through pregnancy) showed significant deficits in multiple outcome measures.
  2. Children of mothers who drank lightly (1-4 drinks per week) had “no differences in test performance,” and “no statistically significant associations.”
  3. Children of mothers who drank moderate amounts (5-8 drinks per week) generally showed no statistically significant effects.
  4. “Only weak and no consistent associations” or “no consistent or significant associations were observed” between binge drinking and children’s attention test scores or measures of executive functions.

To the authors’ credit, they accurately report their findings as a lack of significant associations and then go on to put this finding into context. To cite the abstract of one paper:

“Furthermore, as no safe level of drinking during pregnancy has been established, the most conservative advice for women is not to drink alcohol during pregnancy. However, the present study suggests that small volumes consumed occasionally may not present serious concern.”

Yet, let me list a few of the news articles reporting these results:

This is an example of perhaps the single most common, most stubborn, and frankly most pernicious distortion of research findings about risk.

No evidence of associations DOES NOT EQUAL “SAFE.” Period. End of story.

It’s an ancient logic rule that absence of evidence is not evidence of absence. Just because we do not find an association between alcohol use in pregnancy and measures of child functioning does not mean that moderate or light alcohol use by pregnant women will never harm any children.

What these studies DO show is that such effects, if they exist, must be (a) rare and/or (b) mild enough that well designed research studies were unable to detect them. In other words, the general principle that minimizing alcohol use by pregnant women still holds (as the study authors explicitly note). What we know is that the dose response relationship is such that low doses do not result in detectable effects.

The problem is that we know from decision and risk perception research that people draw sharp distinctions between certainty and uncertainty. 0% chance of harm is unequivocally good. 0.0001% chance of harm — not so much.

Most people equate “safe” with 0% risk. 0.0001% is not “safe” in their minds. So when reports of risk research use that word, as many of the reports about maternal alcohol use yesterday did, they create expectations of absolute safety in the public that are unjustified and unhelpful.

Now, I’m not the only one complaining. In particular, the UK’s National Health Service quickly posted a rebuttal, titled “Daily drinking in pregnancy ‘not safe’.” But far fewer people are likely to see their post than read the many articles that misrepresent this finding.

I try to avoid bashing the media. Most reporters do their best to report research results despite the fact they are not academics and face the difficult task of translating academic language into clear messages for a general public audience. I have a great deal of respect for the challenges they face, especially given today’s 24 hour news cycle and the pervasive general cutbacks in news reporting resources (which lead less scientifically trained reporters to be stuck doing science stories). So, I tolerate a lot of unintentional misrepresentation. But, this is too clear of a problem to let go, and it has way too simple of a solution.

I hereby call for a permanent, universal, unequivocal BAN on presenting anything as “safe” in public reporting and discussion of health risk.

It is never helpful. It denies the possibility of residual risk. It denies the possibility of unanticipated complications. It creates expectations that harm is impossible, and in so doing it undermines the credibility of every single piece of risk research when harm does eventually occur, as it almost always does to someone, somewhere, in some way. Even if that harm was caused by something else or just completely random, people look at the occurrence of any negative event as evidence something is not “safe.”

It is a mistake to label a vaccine, a water supply, or a behavior such as moderate alcohol use in pregnancy as “safe.” At least in today’s world.

When my grandparents were growing up, everyone knew the world was not “safe.” Water supplies were tainted, children got communicable diseases on a regular basis, people changed shirts midday because of air pollution etc. In that context, claims that something was “safe” didn’t mean “perfectly harmless” but simply “a whole lot less harmful than the alternative.”

Today, we have eliminated certain health risks and reduced and mitigated many many more. As a result, people do not expect harms to befall them in everyday life. And, as a result of that change in expectations, we have all redefined what it means for something to be “safe.” The only problem is that the new, everyday definition of “safe” is not the same as what is almost always intended in public health risk communications.

So here is my plea:

To every reporter, editor, news anchor, public health official, university public relations officer, researcher, congressman, community activist, regulator, parent, and educator:

Talk about how research shows that certain behaviors, interventions, or products are safer than others. Talk about keeping people “as safe as possible.” Explain how research studies that find no evidence of an effect can mean that we can probably tone down our worry or stress about those risks, because it is very unlikely that a large and common risk exists and we didn’t notice it.

Just stop calling anything “safe.”


Brian J. Zikmund-Fisher is an Assistant Professor of Health Behavior & Health Education at the University of Michigan School of Public Health and a member of the University of Michigan Risk Science Center and the Center for Bioethics and Social Sciences in Medicine. He specializes in risk communication to inform health and medical decision making.

Previous post:

Next post: