Aim: Accurate biomarkers of the acute respiratory distress syndrome (ARDS) may help risk stratification and management. We assessed the relation between several biomarkers and the severity, course and outcome of late onset ARDS in 101 consecutive critically ill patients with new onset fever. Materials and methods: On study days 0, 1, 2 and 7 we measured angiopoietin-2 (ANG2), pentraxin-3 (PTX3), interleukin-6 (IL-6), procalcitonin (PCT) and midregional proadrenomedullin (proADM). ARDS was defined by the Berlin definition and by the lung injury score (LIS). Results: At baseline, 48% had ARDS according to the Berlin definition and 86% according to the LIS. Baseline markers poorly predicted maximum Berlin categories attained within 7 days, whereas ANG2 best predicted maximum LIS. Depending on the ARDS definition, the day-by-day area under the receiver operating characteristic curves suggested greatest monitoring value for IL-6 and PCT, followed by ANG2. ANG2 and proADM predicted outcome, independently of disease severity. Conclusion: Whereas IL-6 and PCT had some disease monitoring value, ANG2 was the only biomarker capable of both predicting the severity, monitoring the course and predicting the outcome of late onset ARDS in febrile critically ill patients, irrespective of underlying risk factor, thereby yielding the most specific ARDS biomarker among those studied.