Antibiotic use and inflammatory bowel disease: number needed to harm? Authors’ reply

We thank Ludvigsson for his interest in our work and for providing additional insights. As noted, our joint works implicate the role of antibiotics in the development of inflammatory bowel disease (IBD), however, there are limitations as well as unique strengths to our study design which should be noted.^{1 2}

Although confounding may exist, our prospective cohort study design limits this possibility. For example, if an individual is not prescribed an antibiotic course for the first 10 years of follow-up, they will initially contribute at-risk time to the ‘no antibiotic exposure’ group. If, however, they are then prescribed an antibiotic course, the following 1–5 years will contribute time to the ‘antibiotic exposure group.’ At the end of these 5 years, the individual will then return and contribute additional time to the ‘no antibiotic exposure’ group until another course is prescribed. In this way, confounding is reduced, as individuals, along with their inherent clinical characteristics, are able to contribute time to both the ‘antibiotic exposure’ and ‘no antibiotic exposure’ groups.

In addition, although smoking represents a possible confounder in our study, we think this is less likely to explain our results. In particular, tobacco use appears to be a risk factor for the development of Crohn’s disease (CD), however, opposingly, tobacco cessation is a risk factor for the development of ulcerative colitis (UC).³ In our study, however, antibiotic use was associated with the development of both UC and CD, suggesting that tobacco use alone would not explain the positive association observed with both IBD subtypes.

Furthermore, although the possibility of reverse causation exists, we also think this is less likely, particularly with clinically apparent manifestations such a fistulising disease. As such, isolated fistulae, even in the absence of other CD manifestations, will often warrant close evaluation as well as a strong consideration of underlying CD.⁴ Thus, fluroquinolone use as a treatment for symptomatic fistulae may occur in concordance with a diagnosis of CD, limiting the possibility of reverse causality. Further, we obtained similar study results with use of both a 1-year and 2-year lag time from antibiotic exposure, and with UC, also minimising this possibility. Last, the small proportion of individuals with isolated fistulising disease, as well as the persistence of our findings at 5 years postantibiotic exposure additionally limit the possibility of reverse causality.

We also agree that further work needs to be completed to disentangle the attributable risk from antibiotic exposure as compared with the risk from an underlying infection. In our study, however, we see evidence of the potential impact that individual antibiotics may have on the development of IBD. In particular, nitrofurantoin, an antibiotic which has minimal impact on the intestinal microbiome, was not found to be associated with the risk of developing IBD, despite the likely presence of an underlying infection in these patients.⁵ Further, prior data have also suggested higher odds of developing IBD as a result of antibiotic use in the setting of an infection, lending further credence to the idea that antibiotics themselves may play a critical role in the development of IBD.⁶ Future research, however, should control for infection subtype and severity, as well as investigate changes in the intestinal microbiome that may contribute to the development of IBD as a result of subsequent antibiotic use.