Question special

Echoing Roger Fan, thank you to everyone that has made this possible, from NEJM and authors to faculty and colleagues. Thrilled to be here!

ProACT seems one of the most pragmatic trials published recently and definitely within the procalcitonin literature. Everything from outcomes (total antibiotic-days and adverse events associated to withholding therapy) to implementation is geared to simulating a real-world scenario and being relevant to clinicians. Focusing on statistical methodology, why did you favor the use of coprimary variables? Does this make it easier or harder to reject the null-hypothesis, especially when compared to composite outcomes? What about regarding clinical relevance? Furthermore, when assessing superiority (total antibiotic-days) and non-inferiority (safety outcome), is it essential to use an intention-to-treat in the former versus a per-protocol analysis in the latter? How do these differ?