Rationale: The predictive factors of treatment failure for ventilator-associated pneumonia (VAP) caused by Pseudomonas aeruginosa (PA) remain uncertain.
Objectives: To describe PA-VAP recurrence prognosis and to identify associated risk factors in a large cohort of intensive care unit patients with PA-VAP.
Methods: From the multicenter OUTCOMEREA database (1997-2011), PA-VAP onset and recurrence were recorded. All suspected cases of VAP were confirmed by a positive quantitative culture of a respiratory sample. Multidrug-resistant PA strains were defined by the resistance to two antibiotics among piperacillin, ceftazidime, imipenem, colistine, and fluoroquinolones (FQ). An extensively resistant PA was defined by resistance to piperacillin, ceftazidime, imipenem, and FQ. A treatment failure was defined as a PA-VAP recurrence or by the death occurrence.
Measurements and main results: A total of 314 patients presented 393 PA-VAP. Failure occurred for 112 of them, including 79 recurrences. Susceptible, multidrug resistant, and extensively resistant PA represented 53.7%, 32%, and 14.3% of the samples, respectively. Factors associated with treatment failure were age (P = 0.02); presence of at least one chronic illness (P = 0.02); limitation of life support (P = 0.0004); a high Sepsis-Related Organ Failure Assessment score (P < 0.0001); PA bacteremia (P = 0.003); and previous use of FQ before the first PA-VAP (P = 0.0007). The failure risk was not influenced by the strain resistance profile or by the biantibiotic treatment, but decreased in case of VAP treatment that includes FQ (subdistribution hazard ratio, 0.5 [0.3-0.7]; P = 0.0006). However, the strain resistance profile slowed down the intensive care unit discharge hazard (subdistribution hazard ratio, 0.6 [0.4-1.0]; P = 0.048).
Conclusions: Neither resistance profile nor biantibiotic therapy decreased the risk of PA-VAP treatment failure. However, the profile of PA resistance prolonged the length of stay. Better evaluation of the potential benefit of an initial treatment containing FQ requires further randomized trials.