Association between fluoroquinolone resistance and resistance to other antimicrobial agents among Escherichia coli urinary isolates in the outpatient setting: a national cross-sectional study

Abstract

Sir, Fluoroquinolones are among the most widely prescribed antibiotics. Recent reports show that quinolones are prescribed in nearly half of all outpatient urinary tract infection (UTI) visits in the USA, although current guidelines recommend against their use as first-line treatment for uncomplicated UTIs. As a result, the prevalence of quinolone-resistant pathogens, particularly Gram-negative bacteria, is on the rise globally. Horizontally transferable genes (qnr) borne by plasmids encode low-level quinolone resistance in Escherichia coli. Previous studies have found that plasmid-mediated resistance to low concentrations of ciprofloxacin increases the selection of higher ciprofloxacin resistance, by allowing the bacterial population to survive and grow to a level at which secondary mutations to higher resistance can occur. This low-level resistance, conferred by multiresistance plasmids, would not be detected by the usual breakpoints for susceptibility used in microbiology laboratories. By prescribing a quinolone to patients infected with these apparently susceptible isolates, physicians may inadvertently facilitate the selection of higher-level resistance. There are important clinical and public health implications. First, the treatment of infections caused by Gram-negative bacteria containing these resistance elements may be less effective than the treatment of infections caused by bacteria lacking these genes. Second, the biological linking of quinolone resistance with other resistance types that can be carried by plasmids means that overprescribing of quinolones and other antibiotics is likely to further fuel the dissemination of multidrug-resistant bacteria, for which there are limited available treatment options. To inform the practice of evidence-based prescribing, we aimed in the present study to ascertain whether there was an association between fluoroquinolone resistance among urinary E. coli isolates and resistance to other antimicrobial agents. We conducted a cross-sectional analysis of antimicrobial surveillance data from a nationally representative network of microbiological laboratories (The Surveillance Network Database—USA) described in detail elsewhere. The results of all routine susceptibility tests were reported based on CLSI criteria adopted by the facility at the time of testing. Aminoglycoside (gentamicin) susceptibility was chosen as a marker because resistance to it in E. coli is almost always plasmid mediated and plasmids containing qnr and other quinolone resistance genes (such as qepA) commonly contain genes for aminoglycoside resistance. Since clinicians may be more likely to prescribe a quinolone if an isolate is non-susceptible to other treatment options, we included only isolates that were also susceptible to at least trimethoprim/sulfamethoxazole, ampicillin or nitrofurantoin, which are commonly used for treatment of UTIs. Multivariable Poisson regression with robust error variances was used to estimate the relative risk (RR) of fluoroquinolone non-susceptibility. Statistical analyses were performed using Stata v.12.0, with P values ,0.05 considered significant. The study was exempt from Institutional Review Board review at Harvard University. The study included 1651506 outpatient urine E. coli isolates tested for fluoroquinolone (ciprofloxacin or levofloxacin) and aminoglycoside (gentamicin) susceptibility between 2000 and 2010 (Table S1, available as Supplementary data at JAC Online). The majority of isolates in the sample were from female patients (89.5%) and the mean age was 46 years. Fluoroquinolone and gentamicin resistance increased significantly over the sample period, from 3.2% in 2000 to 18.2% in 2010 (P,0.001) and from 2.3% to 7.4% (P,0.001), respectively. After controlling for age, sex, region and year, isolates that were non-susceptible to gentamicin were seven times more likely to be fluoroquinolone resistant than gentamicin-susceptible isolates (adjusted RR: 7.0; 95% CI: 6.5–7.5; P,0.001) (Table 1) that were also susceptible to ampicillin, nitrofurantoin and trimethoprim/sulfamethoxazole. The risk of fluoroquinolone resistance was higher for isolates susceptible to only ampicillin (adjusted RR: 7.8; 95% CI: 7.5–8.1) or trimethoprim/sulfamethoxazole (adjusted RR: 8.6; 95% CI: 8.5 –8.8). Across all five models, age, isolates tested in the south and year were positively associated with fluoroquinolone resistance. Our findings show that the risk of fluoroquinolone-resistant E. coli infection in the outpatient setting is heightened by nonsusceptibility to an aminoglycoside, even amongst isolates that were susceptible to three other antibiotics commonly used to treat UTIs. We also found that there was some geographic variation in risk, with a nearly 50% increased risk of quinolone resistance among isolates from the south, holding other factors constant. There are a number of limitations, including the lack of patient-level data on antimicrobial use. However, our results correspond favourably with similar associations of fluoroquinolone resistance with other resistances reported previously in intensive care patients. One explanation for this association is the widespread circulation of plasmid-encoded quinolone resistance genes. To our knowledge, these findings are the first national-level