Prednisone Reduction in ICU Patients With COPD Exacerbation

The aim of this multicenter, investigator-initiated, prospective, randomized, open-label, non-inferiority study is to evaluate a prednisone prescribing strategy, guided by eosinophil blood count compared to the standard (systematic) administration of corticosteroids, in patients with COPD exacerbation requiring ventilatory support. Patients fulfilling inclusion criteria and consenting to participate in the study, will be randomized through a random table generated electronically, to eosinophil-guided group or to control group. In the eosinophil-guided group, prednisone (1mg/kg/day for up to 5 days or during the hospital stay if less than 5 days) is administered only if the eosinophil count is \>2%. If blood eosinophil count is ≤2%, no corticosteroids are given. In the control group: a treatment based on prednisone at a daily dose of 1 mg/kg will be routinely administered for a maximum of 5 days, or during the hospital stay, if it is less than 5 days. Corticosteroid treatment is taken in the morning in patients with NIV, and through the gastric tube in intubated patients. The hypothesis tested is a non-inferiority of the "eosinophil-guided strategy" compared to the standard strategy, with less exposure to corticosteroids. The primary endpoint is the proportion of unventilated patients at day 6 which is set to 50% in the control group. A pre-specified difference \<10% would be a non-inferiority margin. Secondary endpoints are: Number of ICU days alive without ventilatory support within 28 days after recruitment, length of stay in intensive care Unit, the intubation rate in patients initially under NIV, Mortality in the ICU, Hospital mortality. Safety: New onset of diabetes or worsening of diabetes requiring the start or the increase in insulin therapy, Upper gastrointestinal bleeding (2 g drop of Hb requiring blood transfusion or fibroscopy), Uncontrolled hypertensive crisis requiring the introduction of new antihypertensives, ICU-acquired neuromyopathy, Nosocom