Fluoroquinolones are widely recommended as empirical monotherapy for community-acquired pneumonia. Since 1999, case reports of failure of levofloxacin therapy due to levofloxacin-resistant strains of Streptococcus pneumoniae have started to appear. Most worrying is that, in some cases, levofloxacin resistance has been acquired by pneumococci within days of the initiation of therapy. Because use of current clinical antimicrobial resistance breakpoints fail to identify the majority of S. pneumoniae isolates with only first-step mutations, current treatment guidelines not only may have implications with regard to the ability of surveillance programs to detect emerging resistance but may have therapeutic implications as well.
The emergence of Streptococcus pneumoniae strains with resistance to the β-lactam and macrolide antimicrobials has raised concerns regarding the use of these agents for the treatment of suspected or proven pneumococcal infections . As a result, fluoroquinolones with increased activity against S. pneumoniae, such as levofloxacin, moxifloxacin, and gatifloxacin, are now being recommended and used for the treatment of patients who are at risk for infection with multidrug-resistant strains [1–6]. However, there has been relatively little experience with the use of these agents as monotherapy for large-biomass pneumococcal infections such as community-acquired pneumonia (CAP), compared with the β-lactam and macrolide antibiotics, especially with regard to the potential for the emergence of resistance during therapy.
Here I review the in vitro evidence that fluoroquinolone resistance is increasing and examine a number of case reports of failure of treatment with fluoroquinolones from the medical literature, some of which demonstrate the ability of pneumococci to rapidly develop fluoroquinolone resistance during therapy. I speculate on the mechanism by which S. pneumoniae is able to acquire fluoroquinolone resistance so rapidly and discuss the suitability of the clinical and microbiological MIC breakpoints used to determine fluoroquinolone resistance in the laboratory.
Fluoroquinolone Resistance in Vitro
The emergence of S. pneumoniae strains with of resistance to the fluoroquinolones has been described in Canada, Spain, Hong Kong, Eastern and Central Europe, and, to a lesser extent, the United States. In Canada, Chen et al.  found that the prevalence of ciprofloxacin-resistant pneumococci (MIC of ⩾4 μg/mL) increased overall from 0% in 1993 to 1.7% in 1997–1998 (P = .01) and increased in adults to 3.7% (figure 1). In addition, the degree of resistance also increased. From 1994 to 1998, there was a statistically significant increase in the proportion of isolates with a ciprofloxacin MIC of ⩾32 μg/mL (P = .04).
In Spain, Linares et al.  found an increase in ciprofloxacin-resistant pneumococci from 0.9% during 1991–1992 to 3% during 1997–1998. A similar study performed by Perez-Trallero et al.  between November 1998 and October 1999 found that 7% of S. pneumoniae were resistant to ciprofloxacin. Ho and colleagues [10, 11] documented a marked increase in the overall prevalence of nonsusceptibility to the fluoroquinolones in Hong Kong by comparing results of surveillance studies carried out in 1998 and 2000. Over a 2-year period, the prevalence of levofloxacin nonsusceptibility (MIC of ⩾4 μg/mL) increased from 5.5% to 13.3% among all isolates and from 9.2% to 28.4% among the penicillin-resistant strains.
Nagai et al.  carried out surveillance in 10 Central and Eastern European countries between 1999 and 2000. The overall prevalence of levofloxacin resistance was 0.9%. In a surveillance study in Croatia, Pankuch et al.  studied a total of 585 pneumococcal strains isolated from adults in 22 hospitals from 15 Croatian cities between November 2000 and April 2001. Twenty-one strains (3.6%) were quinolone-nonsusceptible. Quinolone MICs were high for 7 strains with the same serotype (23F) tested, and mutations were found in gyrA, parC, and parE. The remaining 14 strains were more heterogeneous and had mutations only in parC and/or parE. In Northern Ireland, ciprofloxacin resistance was linked to penicillin resistance. Eighteen (42.9%) of 42 penicillin-resistant pneumococci were resistant to ciprofloxacin .
Rates of resistance in the United States have been found to be <2% [15–19]. Doern et al.  reported ciprofloxacin resistance rates of 1.4%. The Active Bacterial Core Surveillance Program of the Centers for Disease Control and Prevention (CDC), which was carried out during 1995–1999, reported rates of nonsusceptibility to levofloxacin of 0.2% . They did not include ciprofloxacin among the agents tested. Most recently, Karlowsky et al.  reported a levofloxacin resistance rate of 0.9% among isolates collected during 2001–2002.
In most countries reporting increasing ciprofloxacin resistance, the resistance has been the result of the emergence of de novo resistance in many different serotypes [7, 9, 20]. However, in Hong Kong, the emergence of resistance has been due to the dissemination of a multiply resistant clone that shares an identical multilocus sequence–typing allelic profile with the globally distributed S. pneumoniae strain Spain23F-1. This fluoroquinolone-resistant variant, designated Hong Kong23F-1, was found to have serotype 14 and 19F variants . As noted, in Croatia, one-third of quinolone-resistant strains were serotype 23F . McGee et al.  characterized a collection of strains obtained from a study in Northern Ireland and from the Alexander Project, an international network that was established in 1992 to monitor the development of antimicrobial resistance to major lower respiratory tract bacterial pathogens. Twenty-nine fluoroquinolone-resistant (ofloxacin MIC of ⩾4 mg/L) isolates of S. pneumoniae were selected from the collection of clinical isolates. Of the 29 fluoroquinolone-resistant strains, 8 (28%) belonged to serotype 23F, 8 (28%) to serotype 9V, and 5 (17%) to serogroup 6. Serotypes 35, 22, 34, 14, and 20 accounted for the remaining 8 isolates.
Studies in the United States have also noted pockets of increased rates of fluoroquinolone resistance. The Prospective Resistant Organism Tracking and Epidemiology for the Ketolide Telithromycin (PROTEKT) study reported that the overall rate of levofloxacin resistance (MIC of ⩾8 μg/mL) in S. pneumoniae was 0.8% (81 of 10,103 isolates) during 2000 and 2001 . The data also indicated that certain states and cities have much higher rates of levofloxacin resistance—for example, 22% of isolates in Salem, Massachusetts—although this may reflect the small sample size of resistant isolates that were identified. A study of 138 S. pneumoniae isolates collected in Brooklyn during 1997 and 1999 showed that ciprofloxacin susceptibility rates decreased from 47% to 16% during this period (P = .0003) . In addition, 5 (3.6%) of 138 isolates were levofloxacin-resistant in 1999.
These in vitro data indicate that fluoroquinolone resistance is increasing and that fluoroquinolone prescribing habits may affect resistance rates. The clinical relevance of in vitro antibiotic resistance has often been questioned, particularly because of the clinical experience with macrolides and β-lactams, which has demonstrated that increasing resistance in vitro does not necessarily result in increases in the rate of clinical failure. Fluoroquinolone resistance appears to be different, however, because cases of treatment failure are appearing.
Clinical Consequences of Fluoroquinolone Resistance
Although treatment failures due to β-lactam, macrolide, and cotrimoxazole resistance in pneumococci have been reported in cases of meningitis and otitis media, the relation between drug resistance and treatment failure among patients with pneumococcal pneumonia is less clear [24, 25]. However, although reports are anecdotal, fluoroquinolone resistance in pneumococci causing pneumonia in association with clinical failure has been well described [26–35]. Reports of the development of resistance and clinical failures appeared shortly after the introduction of ciprofloxacin in 1987 [29, 36–39]. Weiss et al.  described a nosocomial outbreak of infection with fluoroquinolone-resistant pneumococci. During a 20-month period in a hospital respiratory ward where ciprofloxacin was often used as empirical antimicrobial therapy for lower respiratory tract infections, 16 patients with chronic bronchitis developed lower respiratory tract infections caused by a strain of penicillin- and ciprofloxacin-resistant S. pneumoniae (serotype 23F). The ciprofloxacin MIC for all isolates was ⩾4 μg/mL. All 5 patients with acute exacerbations of chronic bronchitis treated with ciprofloxacin experienced therapy failure. Davidson et al.  reported 4 cases of pneumococcal pneumonia treated empirically with oral levofloxacin that resulted in failure of therapy. All cases were associated with the isolation of an organism that either was resistant to levofloxacin prior to therapy or acquired resistance during therapy. Two patients had previously been treated with ciprofloxacin. One of the patients died after 6 days of monotherapy with levofloxacin.
On the basis of these and other studies, a number of risk factors have been recognized that identify patients who are likely to be colonized or infected with fluoroquinolone-resistant pneumococci: patients who are >64 years of age and who have a history of chronic obstructive lung disease and/or prior fluoroquinolone exposure [7, 11, 16, 18, 40]. None of the position papers on CAP published since the introduction of the “respiratory fluoroquinolones” [1–4] has suggested that a history of previous fluoroquinolone use should be a reason for caution when using one of these antimicrobials empirically.
A Model for The Rapid Emergence of Fluoroquinolone Resistance
The case reports described above provide disconcerting evidence that fluoroquinolone-resistant pneumococci can emerge within days of the start of treatment and in patients with no history of fluoroquinolone exposure. To speculate on the mechanism of this rapid development of resistance, we need to examine the acquisition of fluoroquinolone resistance by pneumococci.
Fluoroquinolone resistance in S. pneumoniae is primarily due to mutations in the genes encoding the target topoisomerase enzymes, namely parC, which encodes the A subunit of DNA topoisomerase IV, and/or gyrA, which encodes the A subunit of DNA gyrase . Mutations in parE and gyrB have been reported, but to a lesser extent [32, 42, 43]. The majority of pneumococcal isolates with reduced susceptibilities to fluoroquinolones have amino acid substitutions either in ParC alone or in ParC and GyrA [44–47]. Resistance can also be mediated by active efflux , although its role in contributing to resistance to the newer fluoroquinolones is unclear .
First-step mutations in parC occur fairly frequently: parC mutations occur in ∼1 of 107 pneumococci cultured in the presence of levofloxacin at twice the MIC . This rate, applied to a likely total bacterial load of 1012–1014 pneumococci in an infected lung, means that one could expect 105–107 pneumococci in a patient's lung to have the first-step parC mutation.
Once the S. pneumoniae has a first-step parC mutation, the acquisition of increased fluoroquinolone resistance is dependent on a second-step mutation in gyrA. The development of a first-step mutation appears to facilitate the emergence of a high-level, second-step mutation . Organisms with a first-step parC mutation have an increased likelihood of developing a second-step gyrA mutation, and some estimates suggest that the second-step mutation occurs in ∼1 in 105 first-step mutants [52, 53]. Therefore, the likelihood of a pneumococcus developing a second-step mutation is 1 in 1012, so if there are 1012 to 1014 pneumococci in an infected lung, these data suggest that up to 100 bacteria in that lung would be likely to develop mutations in parC and gyrA and possibly be fluoroquinolone resistant, depending on the inherent activity of the fluoroquinolone in question.
The survival in the lungs of pneumococci showing first- and second-step mutations is obviously dependent on the tissue concentration of the fluoroquinolone that is being used to treat the infection. If this concentration is high enough, first-step mutants will be killed and will be unable to develop the second-step mutation. In contrast, if the fluoroquinolone does not reach the required concentration in the lungs to kill first-step parC mutants, these mutants will survive, multiply, and have the opportunity to develop further mutations and become fully fluoroquinolone-resistant. In this second scenario, there could be a rapid progression to a lung filled with fluoroquinolone-resistant S. pneumoniae.
The Need for Improved Detection of Fluoroquinolone Resistance
The MIC of an antimicrobial is a value that has been used to determine breakpoints that predict the probability of clinical success and/or to detect resistant populations . Clinical breakpoints are dependent on the antimicrobial activity and pharmacologic characteristics of the drug, with the goals of eradication and ultimate clinical success with antimicrobial treatment. In contrast, the goal of microbiological breakpoints is to identify isolates that may be categorized as susceptible on the basis of clinical breakpoints but that harbor resistance mechanisms resulting in reduced susceptibility to the agent being tested. Microbiological breakpoints are therefore useful in monitoring the emergence of resistance. The current NCCLS guidelines make no distinction between these interpretations of the MIC; thus, clinical breakpoints are generally used to characterize resistance to most antimicrobials, including the fluoroquinolones.
Effective surveillance depends on the ability to detect and predict trends in low-level resistance before the development of clinically relevant resistance . However, for the fluoroquinolones, this objective may not be realized with surveillance systems that use clinical as opposed to microbiological breakpoints. For example, the CDC originally used ofloxacin to determine trends of pneumococcal resistance to fluoroquinolones in the United States as part of their Active Bacterial Core Surveillance Program for invasive pneumococcal disease . They were able to document the emergence of ofloxacin-nonsusceptible isolates (MIC of ⩾4 μg/mL), the prevalence of which increased from 2.6% in 1995 to 3.8% in 1997. However, they noted that this was not clinically relevant, because ofloxacin resistance may be seen with a single amino acid substitution in 1 of the topoisomerase targets, whereas the newer fluoroquinolones, such as levofloxacin, require amino acid substitutions in both targets in order for resistance to be detected [18, 56, 57]. To provide more clinically relevant resistance rates, they replaced ofloxacin with levofloxacin as the fluoroquinolone for susceptibility testing in 1998 . It is not surprising, given the increased activity of levofloxacin against S. pneumoniae, that rates of nonsusceptibility to levofloxacin (MIC of ⩾4 μg/mL) were found to be only 0.2% in 1998 and 1999.
Although levofloxacin may be used to predict susceptibility to the other respiratory fluoroquinolones, it fails to detect the emergence of subpopulations of pneumococci with low-level resistance. Lim et al.  found that 59% of isolates with first-step mutations had a levofloxacin MIC of 2 μg/mL, a level that is considered to indicate susceptibility according to NCCLS criteria. Davies et al.  found that of 14 strains with levofloxacin MICs of 2 μg/mL, 10 (71%) had a parC mutation. Lim et al.  also examined strains with levofloxacin MICs of 1 μg/mL and found that 8 (25%) of 32 of these selected few isolates already had first-step mutations. These results indicate that the use of existing MIC breakpoints as a surveillance marker for fluoroquinolone resistance is clearly inadequate. Furthermore, no test is currently available to accurately identify those isolates with low-level resistance, to define these microbiological breakpoints (figure 2) . Even changing surveillance indicators to a different fluoroquinolone, such as ciprofloxacin, does not produce significantly better results: with a ciprofloxacin MIC of ⩾4 μg/mL defining nonsusceptible isolates, 4 (29%) of 14 isolates in the susceptible category harbored first-step mutations.
In addition to underestimating the emergence of fluoroquinolone resistance, failing to recognize isolates with low-level resistance may also result in the use of inappropriate therapy. As noted previously, there is growing evidence that isolates with existing first-step mutations and low-level resistance are more likely to acquire subsequent mutations that result in high-level resistance [51–53]. Therefore, it may not be appropriate to treat a patient with a large-biomass infection, such as pneumococcal pneumonia, with a fluoroquinolone if the infecting isolate has a first-step mutation. The development of high-level fluoroquinolone resistance during therapy has been documented previously in patients with pneumococcal pneumonia .
Fluoroquinolones, such as levofloxacin, are widely recommended for empirical monotherapy for CAP, despite increasing rates of in vitro resistance among S. pneumoniae. Since 1999, case reports of failure of therapy with levofloxacin have started to appear, and, in some cases, levofloxacin resistance has developed within days of the initiation of fluoroquinolone therapy. Fluoroquinolone resistance is primarily mediated through parC and gyrA mutations that affect topoisomerases that control pneumococcal DNA replication. The mechanism by which levofloxacin resistance is rapidly acquired during therapy is open to speculation, but it may be related to the presence of a large number of pneumococci in an infected lung, the relatively low frequency of first-step mutations, and the ability of these mutants to survive therapeutically achievable tissue concentrations of levofloxacin. With the increasing incidence of fluoroquinolone resistance, it may be appropriate to revise the clinical guidelines for empirical fluoroquinolone monotherapy. The microbiological breakpoints used to determine fluoroquinolone susceptibility and the associated clinical breakpoints used to guide treatment decisions also need to be reexamined, because currently available diagnostic tests are unable to identify pneumococci with first-step mutations. In the longer term, however, the introduction of new molecular diagnostic tests for mutant pneumococci may provide a more reliable guide for clinicians.