De-escalation (DE) consists in reappraisal of antimicrobial therapy as soon as microbiological susceptibility data are available. One of its aims is to limit the emergence and spread of multidrug-resistant (MDR) pathogens by reducing broad-spectrum antibiotic use [1].

Up until recently, a precise or consensual definition of DE was missing and DE in real-life practice remained unclear. A consensual definition of DE based on a six-rank classification of beta-lactams according to both their spectrum and their selective pressure on microbiota among beta-lactams that was elaborated using the Delphi method is now available [2].

Using this new consensual definition, we retrospectively evaluated DE effects on antibiotic consumption and on bacterial resistance emergence and assessed their impact on individual clinical outcome in a population of patients with gram-negative bacilli (GNB)-related ventilator-associated pneumonia (VAP) in whom DE was microbiologically possible. We included all adults admitted in two mixed ICUs from the OUTCOMEREA database. Inclusion and exclusion criteria are provided in Fig. S1 and in supplementary material.

DE was defined as a reduction in spectrum and ecological consequences of pivotal beta-lactam within 5 days following treatment initiation and evaluated using the recent six-rank consensual classification of beta-lactams (Table S1) [2]. Patient outcome, antimicrobial consumption and MDR strain carriage were analysed. More information about methods is provided in the supplementary material.

Among the 182 VAP episodes who received empirical adequate therapy (Fig. S1), DE of the pivotal beta-lactams was feasible in 131 (72 %) and actually performed in 70 (38 %). DE occurred during the first 72 h in 61 cases (87 %) but its intensity was as high as microbiologically achievable in only 61 % of cases. Antimicrobial molecules used as empirical therapy are provided in Table S2. Patient comorbidities and illness severities at admission or during the first 48 h were not associated with DE (Table S3). Conversely, Enterobacteriaceae-related VAP predicted DE and medical reason for ICU admission and empiric use if fluoroquinolones as companion-drug prevented DE (Table S4). After adjustment for these predictors, individual outcome did not differ between groups (Table 1). DE did not impact antibiotic- and carbapenem-free days, but increased the number of group 4 molecule-free days (Table 1). We found a marginal but clinically significant reduction in extended spectrum beta-lactamase-producing (ESBL) colonization acquisition (1.4 vs. 8.2 %, p = 0.07) in the DE group but no difference in rate of global MDR strain carriage acquisition (Table 1).

Table 1 Effect of de-escalation on antimicrobial consumption, multiresistant strain carriage acquisition and individual outcomes
Full size table

The main strength of our work is the use of a consensual definition of DE in a very specific population of VAP appropriately treated within 48 h. The limitations of our work are the absence of control of the companion drugs, elapse time of 5 days chosen for DE and lack of power to demonstrate a significant effect on MDR strain carriage given the low rate of MDR acquisitions in our centres. Our results suggest a beneficial effect of pivotal antimicrobial DE on ESBL-PE but not on global multidrug-resistant acquisitions. As suggested before [3], even short-term exposure to empirical carbapenem therapy may have induced resistance. At least, after adjustment for factors associated with DE, this strategy did not seem to be harmful at the individual level.