Longitudinal Trends in Fluoroquinolone Resistance among Enterobacteriaceae Isolates from Inpatients and Outpatients, 1989–2000: Differences in the Emergence and Epidemiology of Resistance across Organisms

Abstract

We conducted a 12-year study to identify and compare trends in annual prevalence of fluoroquinolone (FQ) resistance among Enterobacteriaceae isolates obtained from inpatients and outpatients in our health care system. A total of 46,070 clinical Enterobacteriaceae isolates underwent susceptibility testing. Although there were significant increases in inpatient FQ resistance for all Enterobacteriaceae, FQ resistance trends differed significantly across Enterobacteriaceae (P < .001). For isolates obtained from outpatients, only Escherichia coli and Proteus mirabilis demonstrated significant increases in FQ resistance (P < .001 for each). Trends in outpatient FQ resistance also differed significantly across Enterobacteriaceae (P < .001). There were significant differences between inpatient and outpatient FQ resistance trends for all Enterobacteriaceae except P. mirabilis and Enterobacter cloacae. Although hospital-wide use of certain antibiotics correlated significantly with inpatient FQ resistance, these correlations differed substantially across organisms. Efforts to elucidate the epidemiology of FQ resistance and identify targets for intervention must recognize and account for the variability of FQ resistance across organisms and clinical settings.

Since their introduction ∼15 years ago, the fluoroquinolones (FQs) have become one of the most important classes of agents in the current antimicrobial arsenal [1]. Early predictions suggested that emergence of FQ resistance, particularly among the Enterobacteriaceae, was very unlikely [2]. Subsequent reports noting the emergence of FQ resistance in the Enterobacteriaceae [3] were of great concern, given that these pathogens cause a substantial proportion of serious hospital-acquired infections [4].

Although recent surveillance studies have identified the contemporary prevalence of FQ resistance [4], few longitudinal data exist regarding the evolution of resistance to the FQ antimicrobial class since the initial introduction of this class of antimicrobial drugs. Characterization of these trends and their relationship to patterns of antimicrobial use, comparison of trends in inpatient and outpatient settings, and exploration of variability in these measures across different Enterobacteriaceae are vital to more clearly elucidate the epidemiology of FQ resistance, predict possible future resistance patterns, and highlight potential areas for intervention. To address this, we performed an ecological study to identify longitudinal trends in FQ susceptibility among Enterobacteriaceae isolates obtained from both inpatients and outpatients. We also investigated possible correlations between annual hospital-wide use of specific antibiotics or antibiotic classes and yearly prevalence of FQ resistance in inpatient Enterobacteriaceae isolates.

Methods

Study site. This study was conducted at the Hospital of the University of Pennsylvania (HUP), a 725-bed academic, quaternary care medical center in Philadelphia. All inpatient clinical isolates from HUP are processed and cultured in a central clinical microbiology laboratory. In addition, this laboratory processes and cultures outpatient isolates for the network of outpatient practices affiliated with the University of Pennsylvania Health System (UPHS). The UPHS outpatient practices include both hospital-based clinics and offices and affiliated community practices throughout the 5-county Philadelphia region. Of note, all inpatients at HUP are adults, but ∼5% of UPHS outpatients are persons <18 years old. All Enterobacteriaceae isolated from inpatients and outpatients during the 12-year study period (1989–2000) were identified through records of the clinical microbiology laboratory and were included in the study.

Microbiological methods. We focused our investigation of FQ resistance on the 6 Enterobacteriaceae most commonly isolated at our institution: Escherichia coli, Klebsiella pneumoniae, Proteus mirabilis, Enterobacter cloacae, Enterobacter aerogenes, and Serratia marcescens. All clinical isolates of these organisms collected during the study period (1989–2000) were included in the study and were categorized by the year of collection and by whether they were obtained in an inpatient or an outpatient setting. If the same organism was identified on >1 occasion during a single patient admission, only the first isolate was included.

All clinical specimens obtained during the study period were processed and cultured in a central clinical microbiology laboratory. Enterobacteriaceae were identified by conventional methods and antimicrobial susceptibilities were determined according to established criteria [5]. Before May 1995, the Vitek system (bioMérieux) was the primary method of susceptibility testing. After this time, the laboratory changed to Microscan conventional panels that were read on the MicroScan Walkaway (Dade Behring). Of note, the following agents were used as markers of FQ susceptibility in different time periods throughout the study: ciprofloxacin (1989–1993), ofloxacin (1994–1996), and levofloxacin (1997–2000). An organism was considered FQ resistant if the MIC of ciprofloxacin was ⩾4 µg/mL or the MIC of ofloxacin or levofloxacin was ⩾8 µg/mL.

The hospital-wide use of the following specific antimicrobial agents was documented: vancomycin, late-generation cephalosporins (i.e., ceftazidime, ceftriaxone, and cefepime), fluoroquinolones (i.e., ciprofloxacin, ofloxacin, and levofloxacin), nafcillin, β-lactam/β-lactamase inhibitors (ampicillin/sulbactam, amoxicillin/clavulanate, and ticarcillin/clavulanate), imipenem, trimethoprim-sulfamethoxazole, aminoglycosides (i.e., gentamicin, netilmicin, amikacin, tobramycin), clindamycin, metronidazole, and chloramphenicol. Of note, the primary FQ on the hospital formulary differed during the study period, as follows: ciprofloxacin, 1989–1993; ofloxacin, 1994–1996; levofloxacin, 1997–2000. In each of these time periods, the primary formulary FQ accounted for >90% of all FQ use. Antimicrobial use was described as defined daily doses (DDD) per 1000 patient-days.

Statistical analysis. The annual prevalence of FQ resistance was calculated as the annual percentage of all organisms identified in the clinical microbiology laboratory demonstrating resistance to FQs. This percentage was calculated separately for each organism of interest and separately for inpatient and outpatient isolates of each organism.

To evaluate the trend in the proportion of positive test results over time, the Cochran-Armitage trend test (χ² test for trend) was performed [6]. To determine whether trends in FQ resistance differed across Enterobacteriaceae, we conducted a logistic regression analysis with FQ resistance as the dependent variable and calendar year and organism as independent variables. To determine whether trends in FQ resistance differed between the inpatient and outpatient settings, we conducted a logistic regression analysis with FQ resistance as the dependent variable and calendar year and patient group (i.e., inpatient vs. outpatient) as independent variables. This analysis was conducted separately for each organism of interest.

For each organism, we investigated whether there was a correlation between the yearly prevalence of inpatient FQ resistance and annual hospital-wide use of specific antibiotics or antibiotic classes. A Spearman rank correlation coefficient was calculated to evaluate the relationship between antimicrobial use and annual inpatient prevalence of FQ resistance. For those organisms for which the annual inpatient prevalence of FQ resistance was correlated with >1 antimicrobial agent, a partial correlation coefficient was calculated to identify independent correlations between use of specific antimicrobial agents and prevalence of FQ resistance. For each organism, all antimicrobial agents for which use was noted to be correlated with FQ resistance by the Spearman rank correlation coefficient (P < .15) were included in the partial correlation analysis for that organism.

A significance level of .05 (2-sided) was used for all tests. Statistical analyses were performed using standard programs in Stata software, version 6.0 (Stata), and StatXact software, version 4.0 (Cytel).

Results

A total of 46,070 clinical Enterobacteriaceae isolates underwent susceptibility testing during the 12-year study period (1989–2000). Of these, 19,683 isolates (42.7%) were obtained from inpatients. The average number of inpatient isolates tested per year was as follows: E. coli, 863 (range, 770–1007); K. pneumoniae, 308 (range, 235–457); P. mirabilis, 173 (range, 156–216); E. cloacae, 166 (range, 103–196); E. aerogenes, 68 (range, 35–100); S. marcescens, 63 (range, 43–104). For all of these organisms, there were significant increasing trends in inpatient FQ resistance during the study period (P, by test for trend: E. coli, P < .001; K. pneumoniae, P < .001; P. mirabilis, P < .001; E. cloacae, P = .033; E. aerogenes, P = .009; S. marcescens, P < .001) (figures 1–6). However, the trend in increasing FQ resistance differed significantly among the Enterobacteriaceae studied (P < .001).

Significant increasing trends in inpatient FQ resistance were also noted for each organism when isolates obtained from urine samples and those obtained from nonurine samples were studied separately. The only exceptions were E. aerogenes isolates from urine (P = .902, by test for trend), E. cloacae isolates from nonurine samples (P = .296, by test for trend), and P. mirabilis isolates from nonurine samples (P = .095, by test for trend).

Of 46,070 total isolates, 26,387 (57.3%) were obtained from outpatients. The average number of outpatient isolates tested per year was as follows: E. coli, 1680 (range, 1247–2571); K. pneumoniae, 230 (range, 175–351); P. mirabilis, 181 (range, 106–308); E. cloacae, 56 (range, 32–101); E. aerogenes, 40 (range, 29–67); S. marcescens, 38 (range, 21–56). Of these organisms, only E. coli and P. mirabilis demonstrated significant increasing trends in FQ resistance among outpatients during the study period (P < .001 for each, by test for trend) (figures 1–6). Once again, the trends in FQ resistance among outpatients differed significantly across the Enterobacteriaceae studied (P < .001).

Similar trends in FQ resistance were noted when isolates of each organism from outpatient urine samples and nonurine samples were studied separately. For E. coli, isolates from both urine and nonurine samples demonstrated significant increasing trends in FQ resistance (P < .001 for each, by test for trend). For P. mirabilis, again isolates from both urine and nonurine samples showed significant increasing trends in FQ resistance (P < .001 and P = .042, respectively). In contrast, for other organisms, none of the isolates from urine and nonurine samples obtained from outpatients demonstrated significant increasing trends in FQ resistance.

Longitudinal trends in FQ resistance for E. coli isolates from both inpatients and outpatients were found to be significantly different (P < .001) (figure 1). Similarly, as shown in figure 2, there were also significant differences when comparing the trends in FQ resistance for K. pneumoniae isolates from inpatients and outpatients (P = .002). However, similar analyses for P. mirabilis and E. cloacae revealed no such significant differences when comparing FQ resistance trends for inpatient and outpatient isolates (P = .497 and P = .115, respectively) (figures 3 and 4). Finally, as demonstrated in figures 5 and 6, for both E. aerogenes and S. marcescens, there were significant differences in FQ resistance trends for inpatient and outpatient isolates (P < .001 and P = .003, respectively).

There were 4 antibiotics or antibiotic classes for which use correlated significantly with the annual percentage of inpatient FQ resistance for ⩾1 of the Enterobacteriaceae studied (table 1). However, the agents for which use correlated with FQ resistance differed substantially across organisms. The annual hospital-wide use of FQ and clindamycin correlated with FQ resistance for 3 organisms, and chloramphenicol use correlated with FQ resistance for 4 of the 6 Enterobacteriaceae studied. For E. cloacae and E. aerogenes, there were no antimicrobial agents for which use correlated with FQ resistance (P < .001 and P = .045, respectively, by test for trend). After calculating partial correlation coefficients to control simultaneously for the trends with multiple antibiotics, there remained a borderline significant correlation between FQ resistance and use of both FQs (r = 0.70; P = .081) and chloramphenicol (r = 0.72; P = .066) in E. coli isolates obtained from inpatients. FQ use also retained a borderline significant independent correlation with FQ resistance in K. pneumoniae isolates (r = 0.60; P = .155). For S. marcescens isolates, clindamycin was the only agent whose use was independently correlated with FQ resistance (r = 0.67; P = .036). For P. mirabilis, E. cloacae, and E. aerogenes isolates, there were no antimicrobial agents whose use was independently correlated with FQ resistance.

Discussion

Since the introduction of FQs at our institution 12 years ago, significant increases in FQ resistance among inpatients have occurred for all Enterobacteriaceae. However, increasing trends in FQ resistance among inpatients differed significantly across the Enterobacteriaceae studied. Furthermore, although the annual prevalence of FQ resistance among inpatients was significantly correlated with several antibiotics or antibiotic classes, there were substantial differences in correlations across Enterobacteriaceae. With regard to outpatient Enterobacteriaceae isolates, we noted significant increasing trends in FQ resistance for both E. coli and P. mirabilis. In addition, trends in FQ resistance among outpatients differed significantly across Enterobacteriaceae. Finally, for E. coli, K. pneumoniae, E. aerogenes, and S. marcescens, trends in FQ resistance differed significantly when comparing isolates obtained from inpatients with those obtained from outpatients.

Our demonstration of increasing FQ resistance among Enterobacteriaceae is supported by a recent report from hospitals in England and Wales of increasing FQ resistance in E. coli, Klebsiella species, Enterobacter species, and P. mirabilis isolates recovered from blood samples [7]. If these trends in increasing FQ resistance continue, the future utility of the FQs will become greatly limited. Given the importance of the Enterobacteriaceae as causes of both hospital-acquired and community-acquired infections [4, 8], the role of FQs as empiric and definitive antimicrobial therapy will be severely compromised.

In seeking to identify possible modifiable risk factors for FQ resistance, we noted that the annual prevalence of inpatient FQ resistance was significantly correlated with institution-wide use of several antibiotics. The association between FQ use and FQ resistance for many of the Enterobacteriaceae confirms in vitro evidence of selection of high-level FQ resistance by serial passage of organisms through increasing concentrations of drug [9]. This association has also been noted in recent clinical studies of E. coli and K. pneumoniae infection [10]. Furthermore, it was noted that, for several organisms (e.g., E. coli and K. pneumoniae), increases in FQ resistance among inpatients were most marked in the final years of the study (1996–2000). Whether this reflects a cumulative effect of FQ use or a differential impact of specific FQ agents (i.e., levofloxacin, which was introduced to our formulary in 1996) is unknown.

We also found chloramphenicol use to be correlated with FQ resistance in a majority of the Enterobacteriaceae studied. Of note, in vitro studies have shown that exposure of E. coli to chloramphenicol can, through outer membrane protein changes, select mutants with FQ resistance [11].

The association between clindamycin use and FQ resistance, although demonstrated for 3 of the Enterobacteriaceae studied, is more difficult to explain. It is possible that the antianaerobic activity of clindamycin, by eliminating indigenous gastrointestinal (GI) tract bacterial colonization, might favor subsequent GI colonization with more-resistant bacteria. In fact, volunteer studies have noted that clindamycin use markedly increases the likelihood of GI colonization with FQ-resistant E. coli following FQ use [12].

The substantial differences across Enterobacteriaceae with regard to trends in FQ resistance among inpatients, trends in FQ resistance among outpatients, and correlations between hospital-wide antibiotic use and FQ resistance among inpatients, suggest that these organisms may behave differently with regard to acquisition or expression of resistance determinants. Indeed, it has been noted that organisms with higher intrinsic MICs of FQs require fewer stepwise resistance mutations (and, theoretically, less selective pressure from antibiotic exposure) to exhibit phenotypic resistance [13]. Although the most common Enterobacteriaceae have low intrinsic MICs of FQs in vivo emergence of FQ-resistant mutants in response to FQ exposure has been noted to occur more frequently for Enterobacter species and S. marcescens than for E. coli [9]. In addition, a recent report noted that, unlike any other Enterobacteriaceae analyzed to date, acquisition of FQ resistance in clinical isolates of P. mirabilis frequently involves mutation of gyrB [14]. Additional investigation into whether such variability in acquisition of FQ resistance determinants might help to explain the differences in emergence of FQ resistance across Enterobacteriaceae is warranted.

Our results also suggest that clinical risk factors for resistance may differ across Enterobacteriaceae. Although, to date, studies investigating the clinical epidemiology of FQ resistance in Enterobacteriaceae have focused almost exclusively on E. coli, future clinical studies should compare the epidemiology of FQ resistance across all Enterobacteriaceae. Such comparisons may offer valuable insights into the continued emergence of FQ resistance.

For the majority of Enterobacteriaceae studied, we noted significant differences in FQ resistance trends when comparing isolates of the same organism that were obtained from inpatients with those obtained from outpatients. Past studies have typically noted higher rates of FQ resistance in E. coli isolates obtained from the hospital setting, compared with those obtained from the community [15, 16]. Although this discrepancy may be explained in part by differences in the magnitude or intensity of FQ use in these settings, FQ use is very common in treating both inpatients and outpatients [17, 18]. It is also possible that patients who become colonized with FQ-resistant pathogens while hospitalized may serve to introduce and disseminate these organisms into the community after discharge from the hospital. Future comparisons of the epidemiology of FQ resistance among inpatients with that among outpatients would be of great value in elucidating the factors responsible for the variability in FQ resistance trends across these settings.

There were several potential limitations to our study. The ecological nature of our study, with the lack of individual-level data, limits the extent to which causal inferences can be drawn and makes control of confounding more challenging [19]. In addition, our analysis of correlations between hospital-wide antimicrobial use and annual prevalence of FQ resistance was limited by small sample size, given that we were limited to 12 units of observation (i.e., 12 years of study). Also, because the unavailability of isolates did not permit molecular epidemiological analysis, we were unable to determine whether our results were due to the presence of multiple unrelated strains or to the clonal dissemination of a few strains. Furthermore, our susceptibility testing protocol changed in 1995. However, the most significant increases in FQ resistance have occurred since 1995, and the susceptibility testing protocol has not been modified during this time period. Our study was conducted in a large academic health system, and our results may not reflect those at other, dissimilar institutions. Finally, comparison of FQ resistance trends among inpatients and outpatients may not be valid if these populations are significantly different. Given the ecological nature of this study, limited data are available regarding these populations. However, according to 1990 US census data, the population that serves as the primary service area of HUP is ∼68% African American and 27% white, and 53% of the population is female. In comparison, 52% of the population served by the UPHS network of outpatient practices in 2001 were African American, 48% were white, and 65% were female. The reasonable similarity between the sex and race distribution of the UPHS outpatient network population and that of the census data from HUP's local catchment area reflects that most UPHS outpatient practices are located in close geographic proximity to HUP and draw their patients from this urban region.

In summary, we found that the incidence of FQ resistance among Enterobacteriaceae isolates obtained from inpatient and outpatient populations has increased significantly. These data highlight the importance of identifying strategies to preserve the utility of these agents. Although FQ resistance has increased markedly among the Enterobacteriaceae, significant differences in the emergence of FQ resistance across organisms and clinical settings (i.e., inpatient vs. outpatient) exist. Efforts to elucidate the epidemiology of FQ resistance, predict future FQ resistance patterns, and identify potential targets for intervention must recognize and account for the variability of FQ resistance across organisms and clinical settings.

References