Statistical model for assessing the proportion of subjects with subjective sensitisations in adverse reactions to foods

The outcome from a Double-Blind Placebo-Controlled Food Challenge is often of a subjective natm e and cannot be measured directly. Reactions to placebo challenges are frequently observed, implying that some of the responses in the study are in fact 'false responses'. In order to adjust for these false responses, previous studies have used various methods, including removing subjects from the analysis who reacted to the placebo. Simply ignoring the false responses can lead to misleading estimates for the true proportion of sensitised individuals. This paper outlines two models which can account for these false responses. In the single challenge study, a simple model is developed which enables the estimation of the rate of false responses in the study, as well as the true proportion of sensitised subjects. This model is very easy to apply in practice. For a repeated challenge study, a more complicated model is developed which again enables the estimation of the rate of false responses and the true proportion of sensitised subjects.