De-escalation aims at reducing the use of broad-spectrum antibiotics and therefore the emergence of multidrug-resistant (MDR) pathogens.

Observational studies suggested that this strategy seems to be safe. However, there is no adequate, direct evidence showing de-escalation of antimicrobial agents to be effective and safe for onco-hematology patients with sepsis or septic shock. Thus, randomized clinical trials are needed for testing the safety and efficiency of de-escalation of antimicrobial therapy.

The investigator's hypothesis is that de-escalation of empirical antimicrobial therapy in onco-hematology patients with sepsis or septic shock is noninferior to the continuation of empirical antimicrobial therapy.

The first aim of the study is to demonstrate that de-escalation is noninferior to the continuation of broad-spectrum antibiotics in terms of hospital mortality.

The secondary aims are to compare the two strategies in terms of mortality, duration of antimicrobial therapy, durations of mechanical ventilation, vasopressor use, numbers of superinfections, organ failure.

Antimicrobial de-escalation (ADE) of antimicrobial therapy is a strategy proposed to allow for the rational use of broad-spectrum antimicrobial therapy as the empiric treatment for infections and minimize the overall exposure to these broad-spectrum agents. The need for prompt, effective antimicrobial therapy for patients with known or suspected infections is widely accepted. This principle leads to the use of very broad-spectrum antimicrobial therapy to increase the odds that all suspected potential pathogens are adequately treated. However, the potential drawback is selection of multidrug-resistant (MDR) organisms.

ADE is widely recommended in the management of antimicrobial therapy in intensive care unit (ICU) patients. The Surviving Sepsis Campaign guidelines describe and recommend the process for selecting antimicrobial therapy as commencement of antimicrobials within the first hour, antimicrobial therapy broad enough to cover all likely pathogens, and daily reassessment for potential ADE.

To date, no randomized study assessing this strategy is available for this specific population of cancer critically ill patients. In a recent systematic review based on 13 observational studies and one randomized controlled trial, the authors conclude that the equipoise remains and a large randomized trial is required to assess the effect of the antibiotics de-escalation strategy on the bacterial ecosystem, on MDR carriage, and on patient outcomes.

An interim analysis planned after inclusion of 233 patients.

* Subgroup analyses will be performed on patient subsets:

• Patients with allogeneic hematopoietic stem cell transplant,
• Neutropenic patients (Neutrophils < 0.5 Giga/L),
• Hematological disease,
• Oncological disease,
• Polymicrobial sepsis,
• Multi-drug resistant organisms,
• Patients presenting with bacterial pneumoniae,
• Patients presenting with Intra-abdominal infection,
• Patients presenting with bacteraemia,
• Patients presenting with gram negative bacteria infection,
• Patients presenting with gram positive cocci infection,
• Patients presenting with septic shock,
• Patients presenting with sepsis.

Inclusion Criteria:

1. Written informed consent from the patient or proxy (if present) before inclusion or once possible when patient has been included in a context of emergency

2. Age ≥ 18 years,

3. Onco-hematology patient admitted to intensive care for sepsis or septic shock according to the following criteria:

   • Sepsis:

     • A suspected infection
     • And an acute increase of ≥ 2 SOFA points (a proxy for organ dysfunction)

   • Septic shock:

     • sepsis
     • and vasopressor therapy needed to elevate MAP ≥65 mm Hg and lactate >2 mmol/L despite adequate fluid resuscitation

4. Patient treated with an empirical antibiotic treatment,

5. Patient with at least one microbiological sample collected at least within the first 48 hours following the diagnosis of sepsis in ICU

6. Patient with an identified infectious site according to the definitions,

7. Patient with an identified bacteria microorganism after microbiological examination,

8. Patient affiliated to the national French statutory healthcare insurance system or beneficiary of this regimen.

Exclusion Criteria:

1. Patient colonized with a multi-drug resistant organisms preventing de-escalation antibiotic,
2. Pregnant or breast-feeding woman,
3. No affiliation to the national French statutory healthcare insurance system,
4. Patients deprived of liberty or placed under the authority of a tutor,
5. Inappropriate probabilistic antibiotic treatment,
6. Expected mortality within 48 hours,
7. Patient admitted to the ICU for end-of-life care (do-not-resuscitate patients) . Do-not-intubate (DNI) patients can be included.