Correspondence: Bezlotoxumab for Prevention of Recurrent Clostridium difficile Infection

To the Editor: In the analysis of the effects of a given agent on the recurrence of C. difficile infection, patients need to have been cured of their initial infection before acquiring a recurrent infection.

Furthermore, patients might die during follow-up; thus, death is a classic competing event for cure and recurrent C. difficile infection.1-3 Wilcox et al. censored data from patients without a clinical cure at the date of infusion of the medication. Conditioning on the future violates one principle of time-to-event analyses4; censoring competing events leads to bias (an overestimation of the risk of infection recurrence).

We emphasize that “get cured, stay alive, and remain free of recurrent infection over time” is a more relevant end point for patients with C. difficile infection. Using a multistate model, we have schematically displayed the competing risk bias and reconstructed our proposed end point (Figure 1). Our analysis indicated the possibility that although both active-treatment groups had a lower risk of recurrent infection, the probability of being cured, alive, and free of recurrent infection is lower in the actoxumab–bezlotoxumab group than in the placebo group for the first 5 weeks. Such time-dependent effects are hidden in the original analysis but are highly relevant from the patients’ perspective and should therefore be made transparent.

Harriet Sommer, Dipl.-Math.
University Medical Center Freiburg, Freiburg, Germany
sommer@imbi.uni-freiburg.de

Jean-François Timsit, M.D., Ph.D.
Paris Diderot University, Paris, France

Martin Wolkewitz, Ph.D.
University Medical Center Freiburg, Freiburg, Germany

for the COMBACTE-NET Consortium

DOI: 10.1056/NEJMc1702531