Resistance to third-generation cephalosporins among gram-negative nosocomial pathogens is associated with increased mortality, length of stay, and hospital costs (1–4). Measures to reduce the extent of resistance are therefore warranted.

This study was designed to test the hypothesis that recipients of fluoroquinolones are protected against infection and colonization with the three most common third-generation cephalosporin-resistant gram-negative nosocomial pathogens, Enterobacter spp., P. aeruginosa, and K. pneumoniae (4). We have demonstrated a protective effect of fluoroquinolone use on infection or colonization with these resistant organisms both in crude analysis and after control for confounding variables. Moreover, subgroup analysis demonstrated this protective effect for each genus individually, though small numbers of patients with cultures positive for P. aeruginosa and K. pneumoniae precluded statistical significance in these groups, due to the limited power associated with subgroup analysis. Other notable findings are that surgery, intensive care unit stay, and receipt of a β-lactam/β-lactamase inhibitor combination, a ureidopenicillin, or a third-generation cephalosporin increase the likelihood of recovery of these resistant pathogens. Although we did not match case-patients and controls based on date of admission, division of the entire study period into three time intervals showed the ratio of cases to controls to be approximately the same in each. The likelihood of spurious associations resulting from disparity between the year of hospitalization of cases and controls is therefore minimal.

Our analysis did not differentiate between infection and colonization with the pathogens studied. Since the focus of the study was the occurrence of third-generation cephalosporin-resistant nosocomial organisms in the population we studied, this distinction was not necessary. The organisms we studied are capable of causing infection in a given patient at any point after colonization. Moreover, once they have colonized a patient, they are capable of transmission to other hospitalized patients, in whom they can cause infection. Our objective, then, was not to compare rates of active disease between hospitalized groups, but rather to use the recovery of these organisms as a marker for actual or potential disease in the populations we examined.

In addition to infection control measures, such as active surveillance, hygiene, and isolation precautions, the other important strategy in checking the emergence and spread of antimicrobial resistance is the manipulation of selective antimicrobial pressure through changes in use of antimicrobial drugs (5). Previous studies exploring the effect of antibiotics on third-generation cephalosporin resistance focused on replacement of cephalosporins with other β-lactam–containing agents (8–11). No interventions involving a substitution with a fluoroquinolone have been reported.

Two main categories of β-lactamases mediate resistance to third-generation cephalosporins among the common gram-negative nosocomial pathogens: chromosomal β-lactamases and plasmid-associated extended-spectrum β-lactamases (ESBLs) (12). Enzymes that can confer resistance to most penicillins, cephalosporins, and monobactams, ESBLs belong to Bush-Jacoby-Medeiros functional group 2, whose enzymes are generally inhibited in vitro by β-lactamase inhibitors. By contrast, the chromosomal β-lactamases present in Enterobacter and Pseudomonas (which constitute 90% of the resistant isolates in our study) belong to group 1, whose enzymes are not inhibited by β-lactamase inhibitors (13).

Earlier studies described interventions carried out when plasmid-associated ESBLs were the main mechanism of resistance, so it is not surprising that replacing cephalosporins with a β-lactam/β-lactamase inhibitor combination, as was done in some of these studies (8,9,11), resulted in reduced rates of cephalosporin resistance. Our study, by contrast, found both ureidopenicillins and β-lactam/β-lactamase inhibitor combinations to be risk factors for the isolation of gram-negative organisms resistant to third-generation cephalosporins. We believe that this discrepancy relates to the fact that the predominant cause of resistance in our hospital during the study period was group 1 chromosomal β-lactamases (against which β-lactamase inhibitors are not active) and that plasmid-mediated ESBLs played only a minimal role (14).

Our findings expand on the observations of Kaye et al. regarding the protective effect of fluoroquinolones on the emergence of third-generation cephalosporin-resistant Enterobacter spp (6). They diverge, however, regarding risk factors. Although Kaye et al. found third-generation cephalosporin exposure to be an independent risk factor for emergence of resistance; no other antimicrobial exposure or hospital event was independently associated with this finding. We propose that the suggested discrepancy between these results and our findings that certain hospital events and antimicrobial classes confer enhanced risk for initial isolation of resistant organisms can be attributed to the difference in the design of the two studies.

Kaye et al., in examining emergence of resistance, identified clinical isolates for which β-lactamase production was induced or derepressed mutants were selected. Our study design, by contrast, detected those patients colonized or infected by an organism with preexisting third-generation cephalosporin resistance, a phenomenon made more likely by certain hospital events or antimicrobial drug exposures. Whereas Kaye’s case-patients began with susceptible isolates that developed resistance after a specific exposure, our case-patients were enrolled with already resistant strains. Thus, while a particular hospital event or antimicrobial drug exposure may not induce β-lactamase production or select derepressed mutants, it may well confer enhanced susceptibility to the acquisition of a strain in which resistance mechanisms are already expressed.

Although we do not include molecular typing or epidemiologic data regarding patterns of antimicrobial drug use and colonization with resistant organisms, earlier studies conducted during our study period at the same institution have answered many of these questions (14,15). These studies showed that colonization with ceftazidime-resistant gram-negative bacilli in intensive care units during a nonoutbreak period was common, was probably acquired before admission to the unit, involved diverse strains, and was associated with prior exposure to a variety of β-lactam antimicrobial drugs.

When one generalizes from the interventional studies performed to date, replacing third-generation cephalosporins with other agents is difficult because they are prone to several possible biases: 1) they are before/after studies and are therefore prone to time effect bias; 2) they describe a group-level analysis and are therefore prone to ecologic bias (16); 3) the formulary intervention is usually coupled with improved infection control measures, causing difficulty in determining which measure is responsible for the noted effect; and 4) these studies are more likely to be reported and published when a positive effect is noted, i.e., publication bias.

Our data as well as those of Kaye et al. suggest that fluoroquinolones could be substituted for certain types of β-lactam antimicrobial drugs to prevent the emergence and lower the rates of isolation of the most common third-generation cephalosporin-resistant gram-negative nosocomial pathogens. The potential advantages of adding fluoroquinolones to the armamentarium of agents that can be used to combat third-generation cephalosporin resistance are several: they can be administered orally; they are relatively nontoxic and inexpensive; and they may allow the replacement of earlier generation cephalosporins, receipt of which has previously been identified as a risk factor for isolation of third-generation cephalosporin-resistant gram-negative organisms (15).

A limitation of retrospective analyses is the inability to prove what appear to be causal relationships. Statistical associations are interpreted as risk factors, and inverse associations as protective effects. Proof that fluoroquinolones are in fact protective against the isolation of nosocomial third-generation cephalosporin-resistant gram-negative pathogens, as suggested by the inverse association demonstrated here, will require animal models or prospective interventional studies. Such studies will also be required to determine whether reduced third-generation cephalosporin resistance will come at the cost of increased levels of fluoroquinolone resistance, a phenomenon to which Burke has referred as “squeezing the balloon” (17). Fluoroquinolone resistance, not addressed in our study, occurs primarily by means of chromosomal mutation (18), and resistant mutants could potentially be selected for by increased use of this class of antimicrobial agent. Our data, then, provide the impetus for further studies, including a prospective interventional trial to explore the overall protective efficacy of fluoroquinolones against multiresistant gram-negative pathogens.