Many, many of you wrote me panicked emails earlier this week about news coverage of a new study on antibiotics and infants. The study argues that exposure to antibiotics before the age of two was associated with a higher risk of many, many conditions in childhood. Specifically, the study found increase risks of asthma, hay fever, eczema, celiac disease, being overweight or obese and having ADHD. The rhetoric in the study and media coverage is somewhat alarmist. Many people wrote to me to say their Facebook groups are considering shunning antibiotics.
Before I air my list of grievances with the statistics here (it’s long! get ready!) let me say two general things. First: it is generally agreed that antibiotics can be over-used. When I was a kid every time you went to the doctor with a fever they’d just give you penicillin. We know now that was a mistake. Many childhood illness are viral and antibiotics will not work, and overuse contributes to resistance. Doctors are now taught not to prescribe antibiotics unless they know there is a bacterial infection.
HOWEVER, second: antibiotics are really, really valuable for treating bacterial infections. It is hard to understate the benefits the world has gotten from the invention of penicillin. If your child has a bacterial infection — ear infection, urinary tract infection, etc, etc — they should be treated with antibiotics. Living in the developed world, in this era of basically good health (minus COVID-19), it can be easy to forget that treatable bacterial infections kill a lot of people. It would be a shame for a study like this to push people away from antibiotics they need.
With that intro, let’s dive in.
The basic structure of this paper (link to PDF here) should be familiar to any of you who read this newsletter regularly. It’s an observational study of about 14,000 children. The authors have data on antibiotic exposures before the age of 2 and on health conditions among children as they age. The data is from official medical records, based on prescriptions and diagnoses, so is likely to be well measured. The papers compares their various outcomes — obesity, asthma, ADHD, etc — across children who took antibiotics during their first two years (70% of children) and those who did not. Broadly, their conclusion is that for most of these conditions there is a correlation between antibiotic exposure and worse later health.
In some of the media I read on this they were careful to refer to this link as an “association” and not “causal.” That’s good, because the attempts to address issues of selection in this paper are extremely limited, even relative to most papers in this space. Normally, we’d see authors in studies like this adjust for differences like maternal education, family income, other risk factors (prematurity, low birth weight, etc) in all of their analyses. The primary analyses in this paper do not adjust for anything. In some secondary figures and the Supplemental Appendix (always check it out!) they do show results with some adjustments, but since their measure of antibiotic exposure seems to change between the main results and these additional results it is difficult to see how much the controls matter.
This adjustments for confounding variables is made more complicated by the fact that key variables are missing for a large share of the sample (30% missing birth weight, more than 10% missing maternal education) and have to be imputed. I can’t tell quite how they impute them; they reference a general R package for doing this, but do not give specifics.
On top of this, it’s hard to make heads or tails of some of the patterns in the results which they emphasize. For example: developing atopic dermatitis (eczema) is associated with receiving 3 to 4 antibiotic prescriptions before the age of 2, but not with 5 or more. Exposure to cephalosporins (a particular type of antibiotic) is associated with the largest number of conditions, although penicillin prescriptions are responsible for the effect of celiac disease. Penicillin is also associated with a lower risk of autism, whereas cephalosporins are associated with a higher risk.
The authors also distinguish timing. The effects on ADHD appear for both sexes for prescriptions before 6 months, for neither for prescriptions between 6 and 12 months, and for only girls for prescriptions between 12 and 24 months. I have no idea why this would be. In general, the paper gives only a very limited sense of why we might see these relationships. They discuss the microbiome, which we think is important and is in principle affected by antibiotics, but our understanding of this mechanism is very poor. Certainly nothing we know would suggest these random patterns of links at different age groups.
What do I think is, in fact, driving the results in the paper if it is not causal relationships? One obvious possibility is basic differences across families — income, education, etc — which correlate with health and drive the results. A more precise mechanism is that illness is linked over time. For example: we know that kids who have bronchitis as a complication of viral infections early in life are more likely to show up with asthma later. This would appear here as a link between antibiotics (prescribed for the bronchitis) and asthma, but of course it wouldn’t be the fault of the antibiotics. A third possibility is that this is all capturing diagnosis. Some parents are more likely to seek health care, making their children both more likely to be prescribed antibiotics early in life and diagnosed with health conditions later. Some of these conditions (for example, hay fever or eczema) often go undiagnosed by a doctor because they are mild. This means that differences in diagnosis probability may play a role.
I do not know which of these are driving the results, perhaps it is all three, or yet some other thing. What I am confident about is that one should not take these results as causal. Is it possible that antibiotics affect the microbiome and have some small, longer-term impacts on children? Of course. Nothing in this paper rules that out. But nothing in the paper rules it in, either. I hope I do not have to say this but if your child is ill with a bacterial infection, please give them their antibiotic prescription. And if you have done this in the past, please do not now feel guilty about it.