It hasn’t been a great year for Tylenol, or acetaminophen. A story on the public radio program “This American Life” in September highlighted the dangers, including death, of taking even a slightly higher dose of the drug than directed. A series of lawsuits alleged that Tylenol’s maker, Johnson & Johnson, did not adequately warn people of the risks of liver damage associated with its use. New bottles of Extra Strength Tylenol now feature bright red warning labels on the cap.
Amid all this, two new studies emerged in the last six months citing acetaminophen’s risks for pregnant women and long-term consequences for their children. One study, in the International Journal of Epidemiology, said children born to mothers who took acetaminophen during pregnancy were more likely to have behavior problems and slow motor development at age 3. The other study, in JAMA Pediatrics, cited an increased risk of attention deficit hyperactivity disorder, or ADHD, at age 7. And the effects seemed large: a 40 percent increase in hospital diagnosis of behavior problems, for example.
To say these effects were surprising is an understatement. For years, women were told that the only safe pain reliever during pregnancy was Tylenol. A 2010 review article in Reproductive Toxicology summarized studies covering hundreds of thousands of women and concluded there was no increased risk of birth defects from single-ingredient acetaminophen. Based on the U.S. Food and Drug Administration rating system for drugs in pregnancy, Tylenol is safer than ibuprofen, and much safer than aspirin. And the women in these two new studies were not overdosing on Tylenol. They were just taking it as directed.
In media reports, experts cautioned against over-interpretation; “more data is needed” was a common refrain. But for an individual woman, the decision about whether to take Tylenol or not must be made now, not in another five years when more data is available. This means that it’s crucial to figure out what we actually learned from the data in these two studies.
Both were well run, if not perfect. The first study, on behavior and motor development in children at age 3, used a “sibling-controlled” approach with about 2,900 sibling pairs. Basically, researchers compared two children of the same sex, born to the same mother, with varying exposure to acetaminophen in pregnancy. (An advantage to studying siblings is that researchers don’t have to adjust for racial, socioeconomic, educational or other differences across mothers.) This study found that Tylenol exposure increased the risk of poor motor development, behavior problems and language delays. These effects were most pronounced for the children of women who took Tylenol for 28 days or more during their pregnancies.
The second study, on ADHD and behavior problems at age 7, had a much larger sample size — about 64,000 children — though the authors did not compare across siblings. What makes this study compelling is that the researchers did not rely on parental reports about behavior problems; they linked their data to hospital diagnoses and prescription fills for ADHD medication. Again, they found that exposure to acetaminophen increased both the likelihood of a diagnosis and prescription, and more exposure increased the chances even more. Based on the researchers’ estimates, any exposure to acetaminophen increases the risk of an ADHD prescription by 30 percent.
An unfortunate aspect of the way studies like these report their results is that it’s very difficult to get a sense of the actual size of the effect from the headline numbers. For example, a 30 percent increase in risk on a baseline of 10 percent means an increase from 10 percent to 13 percent. A 30 percent increase in risk on a baseline of 1 percent is an increase from 1 percent to 1.3 percent. In terms of the number of people affected, these are quite different.
Doing the magnitude calculation in these two studies is informative. The effects in the first study are extremely large: The researchers estimated that exposure to acetaminophen for 28 days or more would increase the risk of behavior problems from 6 percent to 10.2 percent. The effects in the second study are much smaller, in large part because the base rates are lower. In this data, 28 in 10,000 children every year will fill a new prescription for ADHD medication without prenatal acetaminophen use. With use, this jumps to 36 in 10,000, an increase of only 8 in 10,000.
The magnitudes alone suggest we should probably be more concerned about the effects seen in the first study than those in the second. But magnitudes alone are not enough: We also need to ask whether these effects are actually due to acetaminophen.
Neither of these studies is a randomized controlled trial; to do that, researchers would have had to tell some of the women in the study to take acetaminophen and others to avoid it. Instead, the researchers just surveyed women about their behavior. Some of the women reported taking acetaminophen and some did not. But we have to wonder: What if the women who took acetaminophen were different in other ways, too? And what if it’s those other differences that explain the different outcomes for their children?
In the paper on ADHD, it’s hard to dismiss this concern. The researchers show that women who took acetaminophen in pregnancy were more likely to have been diagnosed with psychiatric illnesses. Such a diagnosis in a mother clearly links to behavior problems for her children. When researchers controlled for this issue, along with other characteristics of the mother, the effects of Tylenol use declined by about half. This doesn’t necessarily tell us that the mother’s psychiatric illness differences were the problem, but it does tell us that the mother’s characteristics, taken together, mattered a lot. The researchers also considered women who had never had a psychiatric diagnosis. In that sample population, the effects of Tylenol use dropped even more.
There’s a final major concern with both papers: Why did the mothers take the Tylenol in the first place? People typically take painkillers for fevers, headaches, inflammation, etc. It remains possible that these symptoms drove the increased risk of behavior problems in their children, not the Tylenol use. For example, it’s well known that fever during pregnancy is linked to bad long-term outcomes for children. Thus, it may have been the fever that negatively affected the children, not the acetaminophen that was used to treat it.
Both studies attempt to get around this by controlling for the reason the women took the acetaminophen. Researchers in the first study, on motor development at age 3, also cleverly ran the same analysis for ibuprofen and found that ibuprofen exposure in pregnancy didn’t matter as much. Unfortunately, these researchers’ sample size — this is the study that used 2,900 sibling pairs — was too small to show that the differences in effects between ibuprofen and acetaminophen were statistically significant. In the end, it’s difficult to fully dismiss this concern. In an ideal world, we would study women who take Tylenol for no reason whatsoever, but that kind of data is unlikely to exist.
Where does this leave us, other than in desperate need of more data? First, it’s clear the first paper’s findings on behavior and motor development at age 3 are more worrisome. The results in the ADHD paper are small and fragile. Second, the encouraging news even with this second paper is that the researchers found large impacts only from extensive exposure. Women who took acetaminophen occasionally — less than 27 times during their pregnancies — didn’t see these effects.
Finally, it’s worth returning to my observation at the start: These results are surprising. It’s crucial when evaluating studies like these to take into account what we already know. A couple of years ago, physicists published a paper suggesting that some particles could go faster than the speed of light. Other physicists were cautious when discussing this finding, precisely because it was so surprising: Their prior beliefs were so strong that one experiment — even a very carefully and thoughtfully designed one — wasn’t enough to convince them. Later, not surprisingly, this finding turned out to be wrong.
We would do well to approach medical studies in the same way. We have a tremendous amount of evidence that Tylenol is safe in pregnancy, and now a little bit of evidence that maybe there are some risks. We should, perhaps, be more cautious than before, but only a little.