Cefepime-Resistant Pseudomonas aeruginosa

complicates treatment of Pseudomonas aeruginosa infections. To elucidate risk factors for cefepime-resistant P. aeruginosa and determine its association with patient death, we conducted a case–control study in Philadelphia, Pennsylvania. Among 2,529 patients hospitalized during 2001–2006, a total of 213 (8.4%) had cefepime-resistant P. aeruginosa infection. Independent risk factors were prior use of an extended-spectrum cephalosphorin (p<0.001), prior use of an extended-spectrum penicillin (p = 0.005), prior use of a quinolone (p<0.001), and transfer from an outside facility (p = 0.01). Among those hospitalized at least 30 days, mortality rates were higher for those with cefepimeresistant than with cefepime-susceptible P. aeruginosa infection (20.2% vs. 13.2%, p = 0.007). Cefepime-resistant P. aeruginosa was an independent risk factor for death only for patients for whom it could be isolated from blood (p = 0.001). Strategies to counter its emergence should focus on optimizing use of antipseudomonal drugs.

morbidity and mortality rates (4,5). When serious P. aeruginosa infections are suspected, early and appropriate antimicrobial drug therapy is crucial because inadequate drug selection has been associated with increased mortality rates (6,7). Complicating the empiric selection of adequate therapy is the increasing prevalence of antimicrobial drug resistance among P. aeruginosa (8)(9)(10). Even in initially susceptible strains, resistance can rapidly develop during treatment (11)(12)(13).
Cefepime, a fourth-generation cephalosphorin, is one of the few agents remaining that has reliable activity against P. aeruginosa. However, increased prevalence of resistance to cefepime among these organisms has been noted (14)(15)(16)(17)(18). As such, elucidating the epidemiology of cefepime-resistant P. aeruginosa is crucial to ensure that this agent remains a viable therapeutic option. Our goals were to identify risk factors for cefepime-resistant P. aeruginosa infections in the hospital setting and to describe the clinical effects of these infections.

Methods
The study was performed at the Hospital of the University of Pennsylvania (HUP), a 725-bed tertiary-care center, and Penn Presbyterian Medical Center (PPMC), a 344-bed urban community hospital. Each hospital is located in Philadelphia, Pennsylvania, USA, and is part of the University of Pennsylvania Health System. The study was reviewed and approved by the University of Pennsylvania Institutional Review Board.

Participants
To investigate risk factors for cefepime-resistant P. aeruginosa, we conducted a case-control study. We identifi ed study participants through records obtained from the clinical microbiology laboratory at HUP, which performs bacterial cultures on all clinical specimens from HUP and PPMC. All adult patients with a positive P. aeruginosa culture result from January 1, 2001, through December 31, 2006, were eligible for inclusion. Each participant was included only one time; the fi rst positive P. aeruginosa culture identifi ed during the study period was used.
On the basis of our fi rst study goal-identifying risk factors-we designated all participants with a cefepimeresistant P. aeruginosa-positive culture result as casepatients and all participants with a cefepime-susceptible P. aeruginosa culture result as controls. All eligible case-patients and controls were included according to the aforementioned eligibility criteria.

Variables
To assess risk factor variables, we used a comprehensive clinical and administrative University of Pennsylvania health system database, which contains data for all hospitalizations since January 1, 2001, and has been used successfully for similar studies of antimicrobial drug resistance (19)(20)(21). Data elements obtained were age, sex, race, hospital (HUP or PPMC), admission as a transfer from another facility (i.e., outside hospital, longterm care facility, rehabilitation center), location within the hospital at the time of culture (i.e., intensive care unit or not intensive care unit), length of hospital stay before culture, prior admission to HUP or PPMC within the past 30 days, Charlson index (22), and all-patient refi ned-diagnosisrelated group (APR-DRG) classifi cation. The following concurrent conditions were also noted: renal insuffi ciency (serum creatinine level >2.0 mg/dL or requirement for dialysis), malignancy, diabetes, cirrhosis, congestive heart failure, chronic pulmonary disease, immunosuppressive therapy, and HIV infection. These variables were based on International Classifi cation of Diseases, Ninth Revision codes; laboratory data; and pharmacy data.

Drug Susceptibility Profi les
We documented antimicrobial drug susceptibility profi les, anatomic site of cultures, and any co-infections. Drug susceptibilities were conducted and interpreted by a semiautomated system (MicroScan WalkAway System, NC16 panel; Dade Behring, St. Louis, MO, USA) or disk-diffusion susceptibility testing in accordance with the criteria of the Clinical and Laboratory Standards Institute (23). Isolates with MIC = 16 (intermediate) or MIC >32 (resistant) were deemed resistant. A multidrugresistant strain of P. aeruginosa was defi ned as a strain with resistance to >3 antimicrobial drug classes (24). We documented all antimicrobial drug treatment administered during the same inpatient admission for up to 30 days before the positive P. aeruginosa culture. We then categorized the drugs by individual agent, class, and spectrum of activity as follows: aminoglycosides (gentamicin, amikacin), quinolones (levofl oxacin, ciprofl oxacin), extended-spectrum penicillins (piperacillin-tazobactam), extended-spectrum cephalosporins (cefepime, ceftazidime), carbapenems (imipenem, meropenem), anaerobic therapy (amoxicillin/ clavulanate, ampicillin/sulbactam, ceftriaxone, imipenem, meropenem, metronidazole, clindamycin), tetracyclines (doxycylcine), and macrolides (azithromycin, erythromycin) (25). During this study period, cefepime was the primary extended-spectrum cephalosporin used at HUP and PPMC, per formulary guidelines. For multivariable analyses, antimicrobial drugs were categorized by agent or class.

Mortality Rates
To assess the relationship between cefepime-resistant P. aeruginosa and mortality rates, we performed a retrospective cohort study, designating the participants with cefepime-resistant P. aeruginosa as the exposed group and those with cefepime-susceptible P. aeruginosa as the unexposed group. We focused specifi cally on rates for those hospitalized at least 30 days.

Statistical Analyses
We calculated the overall and annual prevalence of cefepime-resistant P. aeruginosa among all isolates identifi ed during the study period. We then evaluated the annual prevalence of cefepime resistance over time by performing the χ 2 test for trend (26).
To assess possible associations between potential risk factors and cefepime-resistant P. aeruginosa, we initially conducted bivariable analyses. Categorical variables were analyzed by using the Fisher exact test, and continuous variables were analyzed by using the Wilcoxon rank-sum test (27). The strength of each association was evaluated by calculating an odds ratio (OR) and a 95% confi dence interval (CI). Multivariable analysis was performed by using forward stepwise multiple logistic regression (28). All variables with p<0.20 on bivariable analyses were considered for inclusion in the multivariable model. Backward stepwise multiple logistic regression was also performed to determine whether identifi cation of risk factors varied with the approach to multivariable analysis. Because of the need to adjust for time at risk when investigating risk factors for antimicrobial drug resistance, we required the "duration of hospitalization prior to culture" variable to remain in the fi nal model (29). We also analyzed the interaction between risk factor variables in the fi nal model. Finally, to focus on those isolates likely to represent clinical infection, as per Centers for Disease Control and Prevention criteria, we repeated the analyses on blood isolates only (30).
To assess the association between cefepime-resistant P. aeruginosa and mortality rates for those hospitalized at least 30 days, we conducted bivariable and multivariable analyses in a similar fashion as for the case-control study. As we did for the case-control study, we repeated the analyses on blood isolates only.
We considered a 2-tailed p<0.05 signifi cant. We used STATA version 10.0 (StataCorp, College Station, TX, USA) to perform the statistical analysis.

Results
During the study period, culture results were positive for P. aeruginosa , and cefepime susceptibility was tested for 2,529 patients.  Figure).
Using bivariate analysis to compare exposures, we found several differences between cefepime-resistant and cefepime-susceptible P. aeruginosa (Table 1). Specifi cally, participants with cefepime-resistant P. aeruginosa were more likely to have received an extended-spectrum cephalosporin, extended-spectrum penicillin, or quinolone. Multivariate analysis indicated that prior use of an extendedspectrum cephalosphorin had the strongest association with cefepime-resistant P. aeruginosa (adjusted OR 2.18, 95% CI 1.57-3.04; p<0.001) ( Table 2). Independently associated with cefepime-resistant P. aeruginosa were prior use of an extended-spectrum penicillin or a quinolone and transfer from an outside facility (Table 2). No substantive differences were found in the fi nal model when analyses were limited to blood isolates.
The overall mortality rate among participants was 13.8% (348/2,529). The mortality rate for participants with cefepime-resistant P. aeruginosa was 20.2% (43/213) and for participants with cefepime-susceptible P. aeruginosa was 13.2% (305/2,316) (relative risk [RR] 1.53, 95% CI 1.15-2.04; p = 0.007). After controlling for signifi cant confounders in the multivariate analysis, cefepime-resistant P. aeruginosa was no longer associated with death (Table  3). However, the association between cefepime-resistant P. aeruginosa and death varied signifi cantly, depending on whether the isolate was from the blood or elsewhere. When analyses were restricted to blood isolates, a signifi cant independent association between cefepime-resistant P. aeruginosa and death was found (RR 15.55, 95% CI 3.10-77.89; p = 0.001] ( Table 4).

Discussion
We found the following to be signifi cant factors independently associated with isolation of a cefepime-resistant P. aeruginosa strain in culture of a clinical sample in the hospital setting: prior use of extended-spectrum cephalosphorins, extended-spectrum penicillins, or fl uoroquinolones; and transfer from an outside facility. We also demonstrated that cefepime-resistant P. aeruginosa was independently associated with increased deaths among patients hospitalized for >30 days but only for those for whom P. aeruginosa was isolated from blood.  Past studies have found an association between use of an antipseudomonal agent and emergence of resistance to that same agent (19,31,32). Past studies have also demonstrated that P. aeruginosa resistance to 1 class of antimicrobial drugs is often associated with resistance to other classes (11,33). The tendency for health careacquired P. aeruginosa to become resistant to drugs from multiple classes is well known, and several molecular mechanisms for its intrinsic and acquired resistance have been suggested (10). Our fi ndings not only suggest that prior treatment with cefepime in itself is associated with subsequent emergence of cefepime-resistant P. aeruginosa but that even prior exposure to certain antipseudomonal agents in other classes is associated. Our results emphasize that to devise strategies that prevent further emergence of cefepime-resistant P. aeruginosa, recent prior use of antipseudomonal agents within the same class and from certain other classes must be recognized. The effect of curtailing use of non-β-lactam agents (i.e., fl uoroquinolones) on cefepime-resistant P. aeruginosa prevalence should be formally assessed.
We also found that transfer from an outside facility was associated with cefepime-resistant P. aeruginosa infection. Because transferred patients potentially came from another hospital or from a long-term care facility, these patients might have been more likely to already be colonized with cefepime-resistant P. aeruginosa at the time of admission. That antimicrobial drug resistance is common in long-term care facilities is well known (34). Further work focusing specifi cally on antimicrobial drug-resistant P. aeruginosa infections in long-term care settings is warranted.
The signifi cant association between cefepime-resistant P. aeruginosa infection and increased mortality rates (limited to those patients with P. aeruginosa bacteremia) may result from the fact that bacteremia is a more serious infection than, for example, a urinary tract infection. Alternatively, these results might be explained by noting that a blood isolate is more likely to represent a true infection than is an isolate from other anatomic sites, where isolates are more likely to represent colonization. Nonetheless, our results emphasize the potential serious effect of cefepimeresistant P. aeruginosa infection and the need for strategies to combat its further emergence.
This study has several potential limitations. The fi rst is the ongoing, and appropriate, debate regarding the selection of the control group for case-control studies investigating the association between prior antimicrobial drug use and resistance. Like Harris and et al., we believe that selection of the control group depends on the study question (29,35). In our study, the main question was "What are the risk factors for cefepime resistance among all clinical isolates of P. aeruginosa in the hospital setting?" Thus, we selected patients with cefepime-susceptible P. aeruginosa infection as controls.
Another potential limitation is selection bias, which is always a concern in case-control studies. We believe that any such bias was minimized by the fact that every patient with a P. aeruginosa isolate was eligible for inclusion. Furthermore, all isolates were identifi ed in the clinical microbiology laboratory at HUP, which processes all inpatient cultures at the participating study sites.
Misclassifi cation bias is also a concern in case-control studies. However, the categorization of case-patients and controls and their exposure status was based entirely on preexisting clinical data from the clinical microbiology laboratory. The antimicrobial drug-susceptibility profi les were determined before study initiation, so determination of case and control status did not infl uence these profi les. Furthermore, case-patients and controls were selected without knowledge of their status regarding risk factors of  interest. Thus, we believe any differential misclassifi cation bias was unlikely. Identifi cation of participants in this study was based solely on clinical cultures. As such, that all of these cultures represented true infection is unlikely. For this reason, we performed additional analyses, focusing only on P. aeruginosa blood isolates because these would be expected to meet Centers for Disease Control and Prevention criteria for infection. The results of these secondary analyses did not differ substantively from the primary analyses investigating risk factors for cefepime-resistant P. aeruginosa. Finally, all patients in this study were admitted to either HUP or PPMC. Thus, our fi ndings can only be generalized to similar academic centers. One must also keep in mind the differing resistance profi les at any given institution.
In conclusion, cefepime-resistant P. aeruginosa will negatively affect clinical outcomes, and strategies to counter its emergence are needed. Recognizing recent prior use of antipseudomonal agents, both within the same class and from certain other classes, is needed for devising successful interventions.