select search filters
roundups & rapid reactions
before the headlines
Fiona fox's blog

antibiotics prescribed to pregnant women and incidence of cerebral palsy and epilepsy in babies

Researchers publishing in the journal PLOS ONE have investigated a potential link between use of antibiotics during pregnancy, and cerebral palsy and epilepsy in offspring. A link has been suggested previously, and the authors suggest that it is macrolide antibiotics which are associated with these illnesses. Roundup comments accompanied this analysis.


Title, Date of Publication & Journal

‘Association between antibiotic prescribing in pregnancy and cerebral palsy or epilepsy in children born at term: A cohort study using The Health Improvement Network’ by Wilhelmine Hadler Meeraus et al.

Wednesday 25 March 2015



Study’s main claims – and are they supported by the data

The paper supports the claim that most antibiotics in pregnancy do not increase the risk of epilepsy or cerebral palsy in the baby.

The paper provides plausible, though not conclusive, support for the claim that erythromycin in pregnancy may increase the risk of those outcomes.  An association is reported, but we don’t know whether the relationship is causal.

(There was no difference in incidence of cerebral palsy or epilepsy between children born from mothers that didn’t take any antibiotics and mothers that did take an antibiotic, while there was a difference within the antibiotic group depending on which antibiotic was taken.  Since the overall effect of antibiotics was no increased incidence, but one type of antibiotic appeared to be associated with increased incidence, it may be thought that this means the other type of antibiotic had a positive effect to ‘even out’ the overall result.  It is possible that penicillins have a small protective effect, but the numbers of prescriptions for macrolides were small in comparison to penicillins, so a protective effect of penicillins is unlikely to be the explanation. If penicillins had much of a protective effect, we’d expect to see it in the overall statistics, which we don’t. Since we don’t, it is more likely that macrolides are associated with an increased risk.  This aspect would have been easier to interpret if hazard ratios for the each of the antibiotic types versus no antibiotics had been given, but they weren’t.)




  • Large sample size (almost 200,000 pregnancies, and over 1000 babies who developed cerebral palsy or epilepsy).
  • Adjustment of results for the most obvious confounders.
  • Support for the results by sensitivity analyses that were consistent with the main analysis (sensitivity analyses are those that show how robust the data are to alternative methods of analysis. In this case, the sensitivity analysis was an analysis of only those women who had a single respiratory tract infection during pregnancy. It’s a sensible analysis, as it helps to control for confounding by indication).
  • Comparison of different antibiotics may (though see below*) have been a good way of controlling for confounding by indication in analysis showing increased risk with erythromycin.


  • Data were only on prescriptions: it is possible that some women prescribed antibiotics didn’t take them.
  • It’s not clear how plausible the hypothesis was a priori for erythromycin. While the Oracle II study found an increased risk of cerebral palsy and epilepsy for the combination of erythromycin and co-amoxiclav, it did not find such a risk for erythromycin alone.
  • The study may have been affected by a “damaged survivors” effect, as the authors acknowledge (see belowǂ).

* The interesting analysis is the comparison between erythromycin and penicillin, which showed an increased risk with erythromycin, even after adjusting for the presence of a “potentially neurologically damaging infection”. This analysis depends very much on how well defined that was. It is possible that erythromycin may be more likely to be used for some infections than others, and if infections that are more likely to be treated with erythromycin are more likely to cause neurological damage, then the effect seen with erythromycin may be a result of confounding by indication. Adjustment for “potentially neurologically damaging infection” should have controlled for that, though it would take someone with greater expertise in infections of pregnancy than I have to comment on whether that was sufficiently robust.

ǂ The authors acknowledge the possibility of a damaged survivors effect. If erythromycin saved some pregnancies that might otherwise have resulted in miscarriage or stillbirth, but with neurological damage, then that would make erythromycin seem harmful when in fact it is saving lives. I do not know how likely it is that the results may have been biased by that. Again, someone with expertise in infections of pregnancy would be able to give a more informed comment than I could on how likely this is to have affected the results.



Confounding by indication: this happens when a disease (in this case, infection) which is treated with a given drug (in this case, an antibiotic) is associated with an adverse outcome, thus giving rise to a spurious association between the drug and the adverse outcome.


Any specific expertise relevant to studied paper (beyond statistical)?

I am not an expert in infections during pregnancy. This paper would benefit from review by an expert in that area, who may be able to comment on whether erythromycin tends to be used for different infections than other antibiotics, whether those infections are likely to be associated with a risk of neurological damage, and also on the probability that a damaged survivors effect was in operation.


Before The Headlines is a service provided to the SMC by volunteer statisticians: members of the Royal Statistical Society (RSS), Statisticians in the Pharmaceutical Industry (PSI) and experienced statisticians in academia and research. A list of contributors, including affiliations, is available here.

in this section

filter Headlines by year

search by tag