Abstract: Biochemical markers of inflammation have been suggested as potentially important prognostic variables for patients with SEPSIS admitted to intensive care. More specifically, the temporal evolution of these biomarkers may indicate how well are patients responding to the therapy and consequently, predictive of a positive or negative outcome. One common approach to evaluate biomarker’s prognostic value consists in taking repeated measurements of the biomarkers (for example every day) for each patient and performing day-by-day comparisons between patients who died and patients that survived. This naïve approach does not provide a fully comprehension of the biomarker evolution over time and how it relates to patient’s outcome. In statistical terms, this problem can be stated as how to model binary outcome (died/survived) given a set of longitudinal measurements of a predictor (biomarker) and possibly adjusting the effect to potential confounders. Although this may seem a fairly common situation, there is very little statistical literature addressing this topic and even less examples in the Medical field that use a different approach rather than the naïve one. In this talk, I will discuss possible statistical strategies for the problem described above, using a real example aiming to evaluate the association between the evolution of the C-reactive protein and survival probability for patients admitted to an intensive care unit with community-acquired SEPSIS.