Volume 7 Supplement 2
Empiric antibiotic therapy for ventilator-associated pneumonia: a systematic review of the evidence
© BioMed Central Ltd 2003
Published: 3 March 2003
Ventilator-associated pneumonia (VAP) is associated with significant morbidity and mortality. Several published guidelines recommend early and broad-spectrum empiric therapy for patients with a suspicion of VAP; however, uncertainty exists regarding the indications for and selection of appropriate agents.
To determine whether evidence exists to support the use of a particular parenteral regimen in the empiric treatment of VAP, we conducted a systematic review of the literature for all randomized controlled trials (RCTs) from 1966 to November 2002. Two authors (MWA and JNH) independently searched MEDLINE, EMBASE, the Cochrane Controlled Trials Register library, and the bibliographies of all reviews, for RCTs. Experts in the field and all authors of trials included in this review were contacted to identify unpublished trials. Data were abstracted and trial quality was assessed independently by the two review groups. Additional information was sought when necessary from the corresponding authors of the retrieved trials. Trials comparing similar antibiotic strategies were pooled, and the risk difference and relative risk were reported for the outcomes of mortality, clinical response, superinfections and adverse events.
In the 33 trials identified, a total of 5049 patients were enrolled; 73.0% were ventilated, 65.1% had microbiological evidence of infection, 17.7% had P. aeruginosa isolated from one or more cultures; and the overall mortality rate was 20.5% (922/4492). No trials with a placebo arm were identified; however, 24 different comparisons were evaluated involving 27 unique antibiotic regimens (13 monotherapies, 14 combination therapies). The sample size of trials ranged from 32 to 400 patients. In every trial the primary objective was to evaluate equivalence as opposed to superiority, and no trials were sufficiently powered to determine a difference of 10% or less between groups. The overall quality of the 33 trials was very low; randomization was concealed in only 10, followup was complete in only eight, an intention-to-treat (ITT) analysis was completed in 18, detection of the outcome was blinded in 10, and a double-blinded strategy was employed in only three. All-cause mortality did not significantly differ between treatment groups for any individual trials or when data was pooled. When evaluating clinical response, a significant benefit was demonstrated with amikacin/ceftazidime versus pefloxacin/ceftazidime, risk difference (RD) = 20% (95% CI, 4–36%), in a single low-quality trial. With pooling of trials a significant benefit was also demonstrated with meropenem versus the combination ceftazidime/amikacin, RD = 13% (4–22%); and aztreonam versus an aminoglycoside, RD = 25% (4–45%).
There is some evidence to suggest that choice of parenteral empiric antibiotic therapy for VAP may influence outcome, although study design limitations preclude strong recommendations regarding the optimal choice of therapy. There is no proven beneficial empiric therapy for which to compare alternate regimens; and no trials have ever evaluated the utility of empiric therapy versus a strategy of delayed directed therapy based on culture results.