I’ve discussed in previous blog posts ways in which confounding by indication can completely change the way observational research is interpreted: it can flip common wisdom about labor induction and cesarean delivery risk on its head, and it can lead to bizarre conversations illustrating a researchers’ blind spots when it comes to discussing topics such as depression and hormonal birth control.
The potential for harm in these examples should be clear: avoiding an induction when it’s the safer route, or skipping a highly effective form of birth control for fear of developing depression. But the final example I want to discuss is my favorite case study because it took years to recognize that maybe it’s okay to relieve the pain of infants and toddlers without putting their lungs at risk a decade later.
(Disclosure: sections of this are adapted from a section in the book I coauthored with Emily Willingham, The Informed Parent: A Science-Based Resource to Your Child’s First Four Years.)
In the 2000s and 2010s, study after study found an unsettling link between children’s use of acetaminophen (the ingredient in Tylenol) and later childhood asthma symptoms. The issue came on my radar around my first child’s birth when my husband showed me a New York Times article about the link. (I looked for the specific article, but there were a half dozen, almost none of which included the context of other research and only one of which addressed the research countering the others.)
In short, multiple studies found children given acetaminophen before age 1 were 2-3 times more likely to develop asthma between ages 5-9. There appeared to be a dose-response relationship too: the more acetaminophen they had, the higher their risk. Several studies concluded “that exposure to paracetamol [what acetaminophen is called in Europe, Australia and New Zealand] might be a risk factor for the development of asthma in childhood.” Other studies began suggesting a similar link with antibiotic use in infancy.
These studies showed a clear pattern. But it took several more years, and many more studies, before researchers started asking the question they should have asked a lot sooner: what if the reason babies needing asthma or acetaminophen was itself a risk factor for asthma? Something like, say, respiratory infections?
Indeed, as researchers revisited data sets, conducted more studies, and did meta-analyses, it emerged that the same children who took more acetaminophen and antibiotics as infants and toddlers were the same ones having more frequent respiratory infections than their peers — and respiratory infections early in life are known risk factors for asthma. And when researchers controlled for respiratory infections, the association between acetaminophen and later asthma risk vanished.
In 2013, a systematic review of 64 studies finding a link between childhood asthma and use of acetaminophen and/or antibiotics in infancy concluded, “The weight of evidence of the collected studies in our review strongly suggests that the association of antibiotics with childhood asthma reflects various forms of bias, the most prominent of which is confounding by indication.” (Researchers also began suggesting confounding by indication biased studies on use of acetaminophen during pregnancy and childhood asthma.)
Yet for years, headlines about these studies may have frightened parents out of using acetaminophen when their child was in pain or had a fever, lest they increase their child’s risk of asthma.
After three case studies, I think I can safely stop beating this drum, but just to underscore the key point: when writing about observational studies that suggest unexpected negative effects from a medication or intervention, consider what else might contribute to those effects — including the very reasons for the medication or intervention in the first place.