RESUMO
BACKGROUND: Acinetobacter baumannii (AB) poses a significant threat to critically ill patients in intensive care units (ICUs). Although an association between antibiotic exposure and resistant AB is reported in the literature, a synthesis of evidence in ICU patients is still lacking. AIM: To summarize the evidence on the association between prior antibiotic exposure and the occurrence of resistant AB in ICU patients. METHODS: Online databases were searched for cohort and case-control studies providing data on the association of interest. Carbapenem/multidrug-resistant AB isolation was compared with non-isolation; carbapenem/multidrug-resistant AB was compared with carbapenem/antibiotic-susceptible AB; and extensively drug-resistant AB isolation was compared with non-isolation. Each comparison was subjected to a restricted maximum likelihood random-effects meta-analysis per antibiotic class, estimating pooled ORs. Stratified meta-analyses were performed by study design, outcome type and association-measure adjustment. FINDINGS: Overall, 25 high-quality studies were retrieved. Meta-analyses showed that carbapenem/multidrug-resistant AB isolation was associated with previous exposure to aminoglycosides, carbapenems, third-generation cephalosporines, glycylcyclines, and nitroimidazoles. Increased risk of isolation of carbapenem/multidrug-resistant AB isolation vs carbapenem/antibiotic-susceptible AB was shown for prior exposure to aminoglycosides, antipseudomonal penicillins, carbapenems, fluoroquinolones, glycopeptides, and penicillins. Third-generation cephalosporin exposure increased the risk of extensively drug-resistant AB isolation vs non-isolation. CONCLUSION: This systematic review clarifies the role of antibiotic use in antibiotic-resistant AB spread in ICUs, although for some antibiotic classes the evidence is still uncertain due to the small number of adjusted analyses, methodological and reporting issues, and limited number of studies. Future studies need to be carried out with standardized methods and appropriate reporting of multivariable models.
