The aim of this study is to elaborate nomograms to optimize amikacine first dosing in critically ill patients, using a population pharmacokinetics model elaborated with multicentric data.

French guidelines recommend for probabilistic therapy to reach an amikacin concentration 1 hour after beginning the infusion ≥ 60 mg/L. This target is rarely achieved in the ICU despite a 30 mg/kg recommended dosage. Using data collected prospectively in critically ill patients of Nîmes (France) (1) and Nantes (France), we will elaborate a population pharmacokinetic model on the non-parametric software Pmetrics and on the parametric software Monolix. We will calculate probability of target attainment of Monte-Carlo simulations, using the non-parametric model. Nomograms to determine optimal first dose of amikacin in critically ill patients, according to a few variables previously identified, will be produced.

Inclusion Criteria:

• patients treated with amikacin for sepsis in one of the participating ICU will be included.

Exclusion Criteria:

• patients with aminoglycoside allergy, history of myasthenia, pregnancy, under guardianship.

Patients treated with amikacin for sepsis ou septic shock in the medical intensive care unit of Nantes,France or in an ICU of Nîmes, France will be recruited.