Bayesian analysis of the ICAT·COVID randomized clinical trial

Abstract

This communication provides new effect measures in the multiplicative scale from the ICAT·COVID randomized clinical trial, obtained through Bayesian statistics. These could not be calculated using the traditional frequentist statistics included in the original publication because the benefits of icatibant (a competitive antagonist of the bradykinin B2 receptors) on top of standard care in patients with COVID-19 pneumonia were such that there were no events in the active group.1 Additive effect measures (eg, risk differences) are the most appropriate measures for identifying the population groups that will benefit most from interventions in presence of interactions acting as effect modifiers.2 However, an aspect that multiplicative measures provide where additive effect measures cannot, is an indication of how many times interventions or exposures increase or decrease disease risk (eg, risk ratio, hazard ratio). Furthermore, multiplicative measures are more commonly used in epidemiology, and are more appropriate for outcome measures with strictly positive values, such as counts and the numerators of incidence rates.