Being able to accurately predict probability of death is important for the intensivist. Serum cytokine levels parallel physiological derangements observed in critically ill patients and are used in commonly applied scoring systems and prediction models. Thus, serum cytokine based prediction models of outcome seem to be reasonable and of great interest. In this issue of Critical Care, Gauglitz and colleagues present their prediction equation for paediatric burn patients with concomitant inhalation injury. They found that IL-10 on admission, or IL-6 and IL-7 five to seven days later, may predict outcome in an excellent way. Increased mortality is observed as serum IL-6 and IL-10 levels increase and serum IL-7 levels decrease. However, the complexity of cytokine kinetics in critically ill patients and the variety of factors capable to affect circulating cytokines even in a subgroup of critically ill patients may affect the validity of the results. Also, serum cytokine based prediction models need to be compared to commonly applied prediction models based on clinical parameters. This will enable identification of the most suitable, accurate, cheapest, and easiest to use model to predict outcome.