© 2003 American Thoracic Society
Prenatal antibiotic exposure and subsequent atopyTo the Editor:McKeever and colleagues present observational data associating prenatal antibiotic exposure with increased risk of asthma and atopic illness in a large U.K. clinical database (1). The search to understand the increasing prevalence of asthma has been helped by such studies, and there is now a body of evidence linking postnatal antibiotic exposure with atopy (2). The biological model that has been used to explain this observation involves disruption of the antigenic influences of the developing gut flora in postnatal life, with a resulting failure of maturation of immune responses. It is harder to understand how antenatal exposure to antibiotics could have similar effects, as the fetal gut remains sterile in utero. Although observational studies can demonstrate associations, they cannot prove causal relationships, and one must look at potentially confounding factors. We feel that there are indeed important potential confounders in this study that need to be considered. First, early childhood exposure to antibiotics has previously been shown to be associated with increased atopy rates in this cohort (3). When this exposure was factored into the analysis, the association with maternal antibiotic exposure weakened and only remained significant if three or more courses of maternal antibiotics were given. Similarly, when differing children's consulting patterns, resulting in ascertainment bias, are factored into the analysis, the odds ratios are reduced. Maternal consulting patterns and pre-pregnancy antibiotic exposure may be of relevance and are not included in this model. It may be that different families show different "antibiotic seeking behavior" independent of consultation rates, and that differing personal and parental expectations of receiving an antibiotic may influence prescribing and diagnosis. Increased maternal antibiotic exposure may therefore be a marker for increased subsequent child antibiotic exposure rather than a causal factor. It would be interesting to see the correlation between maternal and early life child antibiotic exposure—information that should be available in the database. We also feel that socioeconomic status and parental smoking status are potentially confounding factors; social–economic demographics are not recorded in the database and not analyzed, although this potentially could be allowed for by post-code markers. Smoking status is incomplete in this database, and so the allowance made in the analysis may potentially be inadequate. These data are interesting and potentially of great practical importance, but we feel that further studies, possibly involving prospective cohorts, are needed before its conclusions can be uncritically accepted.
University of Aberdeen Aberdeen, United Kingdom REFERENCES
From the Authors:We would like to thank Thomas and Price for their interest in our recent article (1). Although they state the data are interesting, they criticize various aspects of the design of the study, and it is these aspects we would like to address. In reply to their comments, we agree that use of antibiotics as a child is an important confounding factor, and therefore adjusted our analyses appropriately. Thomas and Price are incorrect, however, in assuming that these effects remained significant only for three or more courses of maternal antibiotics. The full results were excluded from the paper for brevity but are now given in the table below. Consulting behavior will clearly be an important potential confounder, and we adjusted our analyses for the child's consultations in the first 6 months of life as a proxy marker of family consulting behavior. Interestingly, there was very little correlation between antibiotics use between mother and children in the same family (r = 0.16, Spearmans). Current smoking status with the General Practice Research Database appears to be accurate (2) and when assessing the impact of environmental tobacco smoke (ETS) on asthma this is most important distinction. Our results for the impact of ETS are similar to others (3). Socioeconomic status is not available through post-code markers in the database, and it is one possible confounder that we were unable to adjust directly for; however, smoking habit and maternal age are proxy markers for social class and addition of these variables to the model did not change the results. Our study design is a prospective cohort as we have taken children from birth and followed them through time using their medical records. We agree that other prospective cohorts are likely to be useful. We used the General Practice Research Database because we wanted to establish a large cohort of children and have access to their medical records quickly and cheaply. Clearly, further evidence is needed to support our findings before they are accepted uncritically, and interestingly, our results are supported by a recent paper in a prospective birth cohort that demonstrated an increased risk for use of antiasthma medication at age 4–5 in children whose mothers used antibiotics during pregnancy (adjusted odds ratio 1.8, 95% confidence interval 1.0–3.1) (4). Further studies are still needed to accept or refute these findings and there are many ongoing cohorts that may be able to incorporate General Practice data into their design to help address these questions.
City Hospital, University of Nottingham Nottingham, United Kingdom REFERENCES
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
