Introduction

Pseudomonas aeruginosa is responsible for a large and growing number of hospital-acquired infections. Each year between 1986 and 2003, the National Nosocomial Infection Surveillance (NNIS) System identified P. aeruginosa among the five most frequently reported pathogens. Although fluoroquinolones (including ciprofloxacin, levofloxacin, gatifloxacin and moxifloxacin) and carbapenems (including imipenem and meropenem) are highly active against P. aeruginosa, resistance to these antimicrobial agents is increasing [4, 8]. In most cases, resistance to fluoroquinolones is mediated by alterations in the drug targets, DNA gyrase and/or topoisomerase IV, via mutations in the genes, gyrA and parC [19]. Resistance to imipenem in P. aeruginosa is most commonly due to reduced uptake resulting from the loss of the outer membrane porin protein, OprD, after mutation in the oprD coding or upstream promoter regions, combined with derepression of the chromosomal β-lactamase [12, 18]. Mutations in the gyr genes and in oprD are unrelated, and clinical strains of P. aeruginosa that are resistant to both fluoroquinolones and imipenem are relatively rare (9%) [16].

A less common mechanism of fluoroquinolone resistance in P. aeruginosa is due to mutation in the nfxC repressor gene, mexT, of the multidrug efflux system MexEF-OprN [10]. MexT has been reported to be responsible for regulating the expression of both the efflux pump and the porin, OprD [10, 13]. Thus, a single mutation in mexT can result in both over-production of the efflux pathway, OprN, and simultaneous under-production of OprD [10], resulting in dually resistant P. aeruginosa. While dual resistance to imipenem and ciprofloxacin has been selected in vitro in a strain of P. aeruginosa grown on ciprofloxacin-containing medium [1], selection did not occur when the strain was grown on medium supplemented with imipenem [21]. These findings support the hypothesis that the use of fluoroquinolones may be a risk factor for the emergence and selection of dual resistance in clinical P. aeruginosa. In order to test this hypothesis in a clinical setting, we conducted a nested case-control study using clinical data and isolates collected during a multi-center trial of antibiotic cycling that included fluoroquinolones and imipenem [15, 25].

Methods

Study population and data collection

The data evaluated in this analysis derive from a multi-center, prospective cohort study designed to assess the potential benefits of scheduled rotation of antibiotics in the medical intensive care units (MICUs) of two large midwestern referral teaching hospitals during 2000–2002 [15, 25]. This study was reviewed and approved by the Human Subjects Research Institutional Review Boards at both participating hospitals and at the Centers for Disease Control and Prevention (CDC).

All patients admitted to the MICU for >2 days were eligible for the cohort. A rectal swab specimen was obtained upon admission, then weekly and finally 2 days after discharge from eligible patients. Specimens were plated onto selective media to detect targeted antibiotic-resistant Gram-negative bacteria, including P. aeruginosa, that were resistant to ciprofloxacin or imipenem, using separate plates containing these antimicrobials [15, 25]. Swabs were also plated to screen for vancomycin-resistant enterococci and Gram-negative bacteria with decreased susceptibility to ceftazidime. Bacterial colonies growing on selective media were identified to the species level and tested for antibiotic susceptibility by standard methods [25]. Results and the interpretation of positive clinical cultures that were obtained by a physician’s order were also recorded.

At the time each rectal surveillance culture was obtained, data were recorded for each antibiotic administered in the prior week to establish antecedent antibiotic exposure. Additional demographic and clinical data collected included age, sex, race, Acute Physiology and Chronic Health Evaluation II (APACHE II) score [9], comorbid conditions, recent surgical procedures (≤28 days) and the use of invasive devices during the incident MICU admission.

Case definition and control selection

A case was a study patient identified to have either a clinical culture or a screening rectal swab culture that grew an isolate of P. aeruginosa resistant to both imipenem and ciprofloxacin. The date of acquisition of the dually resistant P. aeruginosa was considered the index date for cases. Three controls were selected for each case; controls were selected randomly from a subset of study patients not meeting the case-definition but who had a clinical or surveillance culture that grew a potential pathogen, including a Gram-negative, Gram-positive or yeast (Table 1 contains the complete list of organisms). Cultured organisms from either a routine clinical culture ordered by a clinician or those identified from the screening rectal swabs were included. To minimize the confounding effect of severity of illness and age, this subset consisted of study patients who stayed in the MICU as long or longer than the case-patient with the shortest MICU stay, who were at least as old as the youngest case-patient and who had an APACHE II score greater than or equal to the lowest score among cases. The date of acquisition of the first potential pathogen from controls was chosen as the index date for controls.

Table 1 Characteristics and culture information of cases and controls from medical intensive care units, 2000–2002
Full size table

Outer membrane protein analysis and pulsed-field gel electrophoresis (PFGE)

Outer membrane proteins were obtained from dually resistant P. aeruginosa isolates by a standard method [20] and then subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). Protein bands were either stained with Coomassie Blue or transferred to polyvinylidene difluoride membranes (BioRad, Hercules, CA) and then hybridized with murine monoclonal antibodies specific for OprD2 or OprN (gifts from N. Gotoh, Pharmaceutical Research University, Kyoto, Japan) [6, 11]. The development of membranes was carried out with alkaline phosphatase conjugated to an anti-mouse secondary antibody using the AP color reagent kit (BioRad) according to the manufacturer’s instructions.

Chromosomal DNA was extracted from all dually-resistant P. aeruginosa isolates, digested by XbaI and analyzed by PFGE using the Chef-Mapper (BioRad, Fremont, CA) with pulse times ramped linearly from 1 to 25 s over a 21-h period. [14]. Gels were visualized with ethidium bromide, and the results were interpreted using the criteria of Tenover [24]. Isolates were compared with others identified at the same hospital, but no comparison was done between facilities.

Statistical analysis

All analyses were performed using SAS ver. 8.0 for Windows (SAS Institute, Cary, NC). Exposures up until the index dates were aggregated for each study patient and compared between cases and controls. Comparisons were performed by first doing frequency distributions of all covariates by case status to identify those variables showing a significant association with the measured outcome (nosocomial colonization or infection with dually resistant P. aeruginosa isolate), thereby making them eligible for further regression analyses. The continuous variables were all dichotomized according to the distribution median. The continuous variables of APACHE II score and age were examined by tertiles. Odds ratios were determined, and the exposure–outcome relationship between fluoroquinolone use and case status was evaluated for interactions with each covariate. Multivariate logistic regression was then performed using a backwards elimination procedure. The Likelihood Ratio Test was used to determine significance.

Results

A total of 44 cases and 132 controls were studied. Cases and controls had comparable median ages and APACHE II scores (Table 1). Rectal or other surveillance culture were the most common culture sites for both cases (31, 70.5%) and controls (63, 47.7%), although sputum or other respiratory specimens were also common among controls (30, 22.7%). As controls were not limited to patients infected or colonized with P. aeruginosa, a variety of bacterial species were represented in this population (Table 1). Fourteen (11%) controls were colonized or had infections resistant to fluoroquinolones but not imipenem. In contrast, only one control (0.1%) had an isolate that was resistant to imipenem alone.

Case-patients had a longer mean hospital stay than controls (30.5 vs. 18 days, P < 0.001) and a longer mean number of days spent in the MICU (12.0 vs. 7.0 days, P < 0.001). In addition, case-patients were hospitalized in the MICU significantly longer before the index date compared to controls (5.5 vs. 2.0 days, P = 0.003). The case-patients did not differ significantly from the controls in overall distribution of comorbid conditions, but several specific differences were noted (Table 2). Univariate analysis revealed that case-patients were more likely than controls to have received amikacin, vancomycin and imipenem, but not fluoroquinolones (Table 2).

Table 2 Univariate analysis of clinical and demographic characteristics associated with cases and controls
Full size table

In a multivariate logistic regression model, exposure to imipenem, length of stay in the MICU ≥8 days and receipt of a tracheostomy prior to MICU admission remained independent predictors of colonization or infection with a dually resistant strain (Table 3). Fluoroquinolone exposure was not a significant predictor of case status in the final model.

Table 3 Multivariable explanatory modela for colonization or infection with Pseudomonas aeruginosa resistant to both a fluoroquinolone and imipenem
Full size table

Twenty-seven dually resistant isolates from case-patients (61%) were available for additional laboratory evaluation; all of these were from rectal surveillance swab specimens, as the isolates derived from clinical specimens were not saved prospectively. The PFGE analysis of six isolates from the first study hospital identified four unique pulsotypes and two clonally related isolates. Among the 21 isolates from the second study hospital, 11 unique pulsotypes were identified as well as six isolates categorized into clonal group “A”, two isolates categorized into group “B” and two isolates categorized into group “C”. Analysis of outer membrane proteins revealed a loss of OprD in all isolates, while OprN was not detected in any isolate.

Discussion

We found no evidence to support the hypothesis that dual resistance to fluoroquinolones and imipenem in clinical isolates of P. aeruginosa is mediated by mutation in a single regulatory gene (mexT) and selected for by fluoroquinolone exposure in these MICU patients. The epidemiologic data did not identify receipt of a fluoroquinolone as a predictor of dual-resistance. OprN, which is over-produced in laboratory isolates of P. aeruginosa with mexT mutations, was not identified by molecular analysis in any of the 19 unique lineages of dually resistant P. aeruginosa. While mutations in the gyr and parC genes were not investigated in this study, loss of OprD was noted in every case. Therefore, it is likely, that these unlinked mutational events occurred concomitantly, leading to dual resistance without the involvement of efflux pumps. In vitro mutational inactivation of the chromosomal porin gene, oprD, associated with imipenem resistance has been shown to occur at frequencies of 10–7 to 10–9 per cell per generation [23], while the development of resistance to fluoroquinolones in vitro has been reported occurring at a frequency of 10–5 to 10–7 [3].

The interpretation of the epidemiologic and molecular results warrants several considerations. First, 27% of patients received fluoroquinolones during their MICU stay or in the week prior to MICU admission, compared to only 4% who received imipenem. The receipt of fluoroquinolones by a greater proportion of patients, both cases and controls, may have masked any subtle effect this exposure posed as a risk factor for acquisition of P. aeruginosa resistant to both a fluoroquinolone and imipenem. Second, antecedent antibiotic exposure information was available for only a relatively short period (1 week) prior to case ascertainment. Any effect of more distant antibiotic exposure could not be determined [22]. Third, we were unable to explore the potentially important interaction of receiving a fluoroquinolone followed by a dose of imipenem, as only one case-patient received both drugs. Fourth, a limited number of isolates, representing 19 unique strain lineages, were available for molecular analysis. Thus, the pool of unique strains was relatively small. No prior studies on the frequency of mexT mutations in clinical strains have been performed and, ultimately, a larger pool of strains may have identified a low-frequency mutational event. Furthermore, because dually resistant clinical strains were not available for additional laboratory analysis, all of the isolates tested represented colonizing strains, which may have differed from those that caused infections. Fifth, we analyzed fluoroquinolone antibiotics as a class, rather than individually, thereby possibly masking the association of any specific fluoroquinolone with dual resistance. In addition, the time-dependency of certain exposures, such as antimicrobial use, was not assessed; exposures were aggregated from the time of admission until the index date. Finally, since independent resistance mechanisms were involved, the study design may have biased the effect that fluoroquinolone exposure had on case status toward the null, since controls were allowed to be resistant to one or the other class of drug.

Of the other antibiotic exposures assessed in the multivariate model, only the receipt of imipenem was found to be associated with the acquisition of a dually resistant P. aeruginosa strain. In addition to imipenem exposure, length of stay in the MICU and presence of a tracheostomy at the time of MICU admission showed significant association with case status. Neither of these findings is surprising. Length of stay has been implicated as a risk factor in numerous previous studies of colonization and infection with multidrug-resistant bacteria [2, 4, 8]. Other studies have associated the presence of invasive devices, such as tracheostomies, with P. aeruginosa infection, and P. aeruginosa is a frequently isolated human respiratory pathogen [7, 17].

In summary, we found that receipt of fluoroquinolones was not a risk factor for nosocomial colonization or infection with isolates of P. aeruginosa that were resistant to both fluoroquinolones and imipenem among MICU patients. This finding, together with the failure to detect OprN in any of the isolates, suggests that the in vitro phenomenon of a mexT mutation and subsequent dual resistance to fluoroquinolones and imipenem occurred rarely, if at all, in these MICU patients. Nevertheless, exposure to fluoroquinolones has been shown to be a risk factor for acquisition of fluoroquinolone-resistant P. aeruginosa and other antibiotic-resistant pathogens [16]. We support prudent use of all antibiotics, including fluoroquinolones and imipenem, as a means of decreasing selective pressure for emergence of antibiotic resistance.