Relationships Between Susceptibility of Pseudomonas aeruginosa and Hospital- and Patient-Specific Variables: Report from the Antimicrobial Resistance Rate Epidemiology Study Team (ARREST Program)

Authors: Bhavnani SM, Hammel J, Forrest A, Ambrose PG, Rubino CM, Jones RN
Conference: ECCMID
Keywords: anti-infectives, antibiotic, cefepime, censored data, ciprofloxacin, correlation, general linear modeling, infectious disease, MIC, minimum inhibitory concentration, P.aeruginosa, piperacillin, susceptibility, tazobactam, tree-based modeling
Division: Cognigen

Abstract

Introduction: Identification of patients with infection associated with antibiotic-resistant pathogens remains a serious challenge for the study of drug regimens to treat such infections. The ARREST Program was established as a multidisciplinary, collaborative effort to use surveillance data and analytic techniques to better understand factors associated with antimicrobial resistance. The analyses presented herein were conducted to identify factors predictive of decreased susceptibility of Pseudomonas aeruginosa in hospitalized patients.

Methods: Five years (1997-2001) of North American SENTRY Program data were analyzed. MIC for cefepime (CPM), ciprofloxacin (CIP) and piperacillin/tazobactam (P/T) vs. patient-specific (e.g., age, hospital stay prior to isolate collection (hospital duration), infection source, specimen, primary diagnosis) and hospital-specific (e.g., bed count, geographical region, study year) variables were analyzed using multivariable general linear modeling for censored data with backwards stepwise elimination (at p > 0.1).

Results: MIC50, MIC range, and % non-susceptible (NS) for isolates (n=487, 93% blood, from 33 hospitals) were: 2, 0.5 to > 16, 14 for CPM; ≤ 0.25, ≤ 0.03 to > 2, 15 for CIP; and 4, ≤ 0.5 to > 64, 26 for P/T. Highly significant variables and interactions between variables identified from multivariable models included hospital duration (p = 0.008) and specimen (p = 0.003) for CPM; specimen (p < 0.0001) for CIP; and hospital duration*primary diagnosis (p ≤ 0.008) for P/T, with higher MICs resulting from combinations of these and other significant variables. Observed MIC50 (% NS) were compared in selected patient cohorts with such combinations (see table). For the patient cohort with at least 2 of the identified characteristics predictive of higher MIC, MIC50 remained stable for each agent while % NS increased markedly for P/T.

European Congress of Clinical Microbiology and Infectious Disease (ECCMID), Glasgow, Scotland, May 2003

By SM Bhavnani, JP Hammel, A Forrest, PG Ambrose, CM Rubino, RN Jones