Introduction

Escherichia coli is responsible for the vast majority of febrile urinary tract infections (FUTI) in men [1]. Although the clinical diagnosis of FUTI in males is relatively easy it is often difficult to establish the exact topographic location of the infection in the urinary tract. It has been demonstrated, with different diagnostic methods such as the measurement of serum prostate-specific antigen (PSA) [1] and the use of transrectal ultrasonography [2] and leukocyte scintography [3], that the prostate is the organ most frequently involved in males with FUTI. Thus, it is considered that most of these infections are in fact acute bacterial prostatitis (ABP) which correspond to the type I prostatitis in the NIH classification [4].

Quinolones, mainly fluoroquinolones (FQ), reach high concentrations in the prostatic tissue and in the seminal fluid and, therefore, are considered to be the first therapeutic choice in males with FUTI, except when resistance to these agents is confirmed or strongly suspected [5]. The widespread use of FQ in veterinary medicine and in the treatment of several human infectious diseases, mainly urinary tract infections (UTI), has led to an increasing resistance to these agents among E. coli isolates [6]. According to the European Antimicrobial Resistance Surveillance Network (EARS-Net) quinolone resistance (QR) has constantly been rising in Europe and this tendency is particularly relevant in Spain in which, as stated in their 2009 surveillance report, 30% of all invasive E. coli isolates are resistant to FQ, which limits empirical treatment with these antimicrobials [7].

Several studies have globally evaluated, both in women and males, risk factors for QR in E. coli strains isolated from patients with community acquired [8–10] or nosocomial UTI [11] but, to our knowledge, none has specifically evaluated this issue in male patients with FUTI. In the setting of ABP, Etienne et al. found that 15% of the E. coli strains isolated from community acquired ABP and 36% of those strains isolated from patients with nosocomial ABP were resistant to FQ, although risk factors for QR were not assessed [12]. As quinolones are the optimal treatment of FUTI in men, identification of risk factors for QR would improve the selection of which males with FUTI could receive quinolones as empirical treatment. Recently, the importance of healthcare-associated (HA)-UTI has been recognised when classifying community infections. In a study by Khawcharoenporn et al. performed among patients with UTI diagnosed in an emergency department (ED) that included both women and males with either cystitis, pyelonephritis or urosepsis, the authors reported that those with HA-UTI had higher rates of levofloxacin resistance than patients with community-acquired (CA)-UTI [13]. The importance of HA-UTI as a risk factor for QR among E. coli strains isolated from males with FUTI has not been, to our knowledge, evaluated.

The objective of our study was to evaluate the prevalence and the clinical factors associated with QR among E. coli strains isolated from males with FUTI.

Materials and methods

This was an ambispective (retrospective and prospective) cross-sectional study in which we collected data from all male patients aged ≥ 18 years with a community FUTI caused by E. coli that attended the ED of our 165-bed primary care hospital from January 2008 to January 2011. Data were recorded retrospectively from January 2008 to October 2009 and prospectively from October 2009 to January 2011. FUTI was defined as an armpit temperature ≥ 38°C and one or more symptoms of UTI (urinary urgency, frequency and/or dysuria), in absence of other sources of infection. Patients with nosocomial infections, defined as infections that were noted 48 h after admission, were excluded from the study. The study was conducted with the approval of the hospital Ethics Committee.

Based on the previous accepted definitions of healthcare-associated bloodstream infections [14], HA-UTI was defined as those UTI affecting patients that fulfilled any of the following criteria: hospitalization in an acute care hospital for 2 or more days in the previous 90 days; residence in a nursing home or in a long-term care facility; received intravenous therapy at home or in a day hospital, hemodialysis treatment, intravenous chemotherapy, 30 days before the infection; received wound care or specialized nursing care in the preceding 30 days. We also included as HA-UTI those FUTI that occurred in patients with long-term indwelling urethral catheters or who had gone through an invasive urinary tract procedure in the previous 30 days. Patients who did not fit the criteria for HA-UTI were considered to have a CA-UTI.

A structured questionnaire including demographic and clinical data was completed for each patient. Clinical variables recorded were: age, dementia, HA-UTI, diabetes mellitus, chronic kidney failure (creatinine concentration > 1.2 mg/dl), cirrhosis, active neoplastic disease (solid tumor or hematologic malignancy diagnosed or treated in the past 5 years), chronic obstructive lung disease, chronic heart failure, use of any immunosuppressive agent including corticosteroids within 30 days of the FUTI, the Charlson comorbidity index [15] and antibiotic treatment prior to presentation of the UTI in the previous 30 days. We also recorded whether the patient had urinary tract abnormalities or previous UTI. Urinary tract abnormalities included: prostate benign hypertrophy or cancer, bladder diverticuli, urethral strictures, congenital abnormalities, renal lithiasis and cysts, neurological bladder and vesicoureteral reflux.

Urine samples were obtained from clean-catch midstream urine or from urinary catheters and cultured on MacConkey agar. Positive urine cultures were defined by bacterial growth ≥ 103 colony forming units/mL [16]. Identification of E. coli was performed by means of standard methods. Patients with polymicrobial urine cultures were excluded from the study. Susceptibilities to amoxicillin, amoxicillin-clavulanic acid, cefuroxime, gentamicin, pipemidic acid, ciprofloxacin, trimethoprim-sulfamethoxazole (co-trimoxazole) and fosfomycin were tested by agar disk diffusion method according to the Clinical Laboratory Standards Institute criteria (CLSI) [17]. Intermediate and resistant E. coli strains to either of the antimicrobials tested were grouped together for data analysis. E. coli were catalogued as an extended-spectrum β lactamase (ESBL) producing strains in case of resistance to ceftriaxone, ceftazidime and/or to aztreonam, susceptibility to cefoxitin and evidence of synergism between amoxicillin-clavulanic acid and third generation cephalosporins, as recommended by the CLSI [17]. PSA levels were measured by monoclonal fluoro-immunoassay (normal PSA level, ≤ 4 ng/ml) in a subgroup of patients. This subgroup of patients did not include those with known active prostatic cancer.

Statistical analysis

Qualitative data are expressed as percentages. Quantitative data are expressed as mean (± standard deviations). Continuous variables are compared by the Student t test or the Mann-Whitney U test when the distribution departed from normality. Categorical data were compared by using the chi-square or the Fisher exact test as appropriate. Any risk factor for QR with a borderline significant P value in the univariate analysis (P ≤ 0.1) was included in a binary logistic regression analysis with QR as the dependent variable. Statistical significance was defined as a two-tailed P value of ≤ 0.05. Statistical analysis was carried out by the program SPSS (version 15.0; SPSS, Inc., Chicago, IL).

Results

During the study period, 153 patients with a community FUTI due to E. coli were attended in our ED. One hundred forty one (92.1%) of the E. coli strains were isolated from midstream urine and 12 (7.9%) from urinary catheters. Table 1 shows the clinical characteristics of the patients included in the study. One hundred twenty (78.5%) patients had a CA-UTI and 33 (21.5%) a HA-UTI. Sixty-six patients (43.1%) were elderly (≥ 65 years) and 68 (44.4%) had one or more associated long-term medical conditions. Among the patients with known urinary abnormalities, prostate benign hypertrophy was the most common and was present in 60 (39.2%) patients. Regarding physical examination, flank pain was present in 35/145 (24.1%) and prostate tenderness in 34/80 (42.5%) of the patients in which either lumbar or rectal examination were evaluated. Additionally, in 79 patients on whom both exploratory maneuvers were performed, flank pain and prostatic tenderness were simultaneously present in 5 (6.3%) cases and absent in 35 (44.3%) patients.

Table 1 Characteristics of the patients included in the study and univariate analysis of risk factors associated with community febrile urinary tract infection caused by quinolone resistant E. coli strains
Full size table

The resistance rates of the E. coli strains included in the study are shown in Table 2. Among the 153 E. coli isolates, 18 (11.7%) were resistant to pipemidic acid and 34 (22.2%) were also resistant to ciprofloxacin. Thus, 101 (66%) of the E. coli strains were catalogued as quinolone susceptible (QS) and 52 (34%) as QR. Table 2 also shows the univariate analysis of the antimicrobial resistance rate differences between QS E. coli and QR E. coli strains. QR E. coli strains were more frequently resistant to other antibiotics including amoxicillin-clavulanic acid, cefuroxime, gentamicin and co-trimoxazole. A higher frequency of ESBL was also observed in the QR E. coli strains.

Table 2 Resistance rates found in the E. coli strains included in the study and existing differences between quinolone susceptible and quinolone resistant E. coli strains
Full size table

Bacteremia was found in 23/103 (22.3%) patients from whom blood cultures were drawn. Among these 103 patients, bacteremia was present in 11/64 (17.2%) of the FUTI caused by QS E. coli strains and in 12/39 (30.8%) of the episodes caused by QR E. coli strains (P = 0.1). Total PSA values were evaluated in 47 (30.7%) of the patients, with a mean value 18.92 ± 29.86 ng/ml, and were elevated (≥ 4 ng/ml) in 31 (65.9%). Although FUTI episodes caused by QS E. coli strains had a higher degree of prostatic involvement (PSA levels 24.7 ± 39.4 ng/ml) compared to those caused by QR E. coli strains (PSA levels 14.2 ± 18.6 ng/ml), these differences were not statistically significant (P = 0.23).

In the univariate analysis (Table 1) males with FUTI caused by QR E. coli strains were more likely to be older, to have a HA-UTI, dementia and higher Charlson scores. They also had a higher frequency of previous UTI, urinary tract abnormalities and antibiotic use in the previous month compared to their E. coli QS counterparts. We performed three multivariate models, as similarly done by others [18]. In each model we included those variables with a P value ≤ 0.1 in the univariate analysis (Table 3). In the first (general) model we included HA-UTI and previous antibiotic use but not their individual components, such as the different forms of health care contacts or specific antimicrobials. The independent risk factors for a QR E. coli FUTI selected in this model were to have a HA-UTI and a previous antimicrobial use. In the second model we introduced the different components included in the HA-UTI definition with a P value ≤ 0.1 in the univariate analysis, as shown in Table 1 (hospital admission, nursing home residence, indwelling urethral catheter, invasive urinary instrumentation), instead of HA-UTI, and the same variables as in the first (general) model. In this model, none of the variables included in the definition of HA-UTI that resulted as significant in the univariate analysis reached statistical significance in the multivariate model. In the third model we introduced FQ use, as this was the only antimicrobial with a P value ≤ 0.1 in the univariate analysis, and the same variables as in the first (general) model. In this final model, prior treatment with FQ was associated with QR E. coli FUTI.

Table 3 Multivariate analysis of risk factors for community febrile urinary tract infections in males caused by quinolone resistant E. coli strains
Full size table

Regarding the antimicrobial empirical treatment, 77 (50.3%) patients were treated with cephalosporins, which represented the most common therapeutic option, followed by FQ in 46 (30%) patients and amoxicillin-clavulanic acid in 21 (13.7%). Only 7 (15.2%) out of the 46 patients that were treated with FQ received an inadequate treatment as the FUTI was caused by a QR E. coli strain. The mean duration of the treatment in the 153 patients included in the study was 16.7 ± 6.8 days, without significant differences between FUTI episodes caused by QS or QR E. coli strains (data not shown).

When analyzing the destination of the patients, 57 (37.2%) were discharged from the ED, 32 (20.9%) were admitted to the observation unit (OU) of the ED, 62 (40.5%) were admitted to hospital, either to the internal medicine or the urology department and 2 (1.3%) were transferred to a tertiary care hospital. The mean hospital stay was 0.6 ± 0.7 days for patients admitted to the OU and 4.6 ± 2.7 days for those admitted to hospital. Those FUTI caused by QS E. coli strains required less days of hospitalization (1.6 ± 2.6 days) compared to those episodes related to QR E. coli strains (2.7 ± 2.9 days)(P = 0.027). Two patients died, both with an QS E. coli urinary infection, one directly related to the infection (septic shock) and another one due to progression of a prostate cancer.

Discussion

In our study, 34% of the E. coli strains isolated from men with community FUTI were QR, the prevalence of which is significantly higher than previously reported by others in the literature although similar to the frequency of QR seen among invasive E. coli isolates reported in the 2009 EARS-Net surveillance report [7]. In Spain, Velasco et al. observed that 10% of E. coli strains responsible for community FUTI in males were QR [19] while Alos et al. found that 25% of the E. coli isolated from men with community UTI were QR [20]. In a study from patients with ABP, Etienne et al. evidenced that 15% of the E. coli strains causing community-ABP and 36% of those isolated from patients with nosocomial ABP were resistant to FQ [12]. On the other hand, Koijjers et al. have recently reported that only 3% of E. coli strains from men with UTI were QR, although patients with fever, urological complaints or with urinary catheters were excluded from this study [21].

In the present study, despite several variables were more prevalent in patients with an infection related to a QR E. coli strain, in the multivariate analysis, only a healthcare related acquisition of the UTI and the use of FQ in the preceding month remained associated to QR E. coli. The relationship between HA-UTI and QR has also been recently evidenced in a study performed by Khawcharoenporn et al. among patients with a community UTI attended in an ED. The authors observed that HA-UTI were more frequently caused by uropathogens resistant to FQ than CA-UTI [13]. In this study 10% of the E. coli strains isolated from CA-UTI and 39% of those isolated from patients with a HA-UTI were QR, frequencies which are very similar to the ones observed in our study. At present, one of the main problems when analyzing HA-UTI is that its diagnostic criteria are not properly defined. In 2002, Friedman et al. demonstrated that health associated bloodstream infections in adults were similar to nosocomial infections regarding the frequency of various comorbid conditions, source of infection, pathogens involved and their susceptibility pattern [14]. Since then, the definition of HA-bloodstream infection used by Friedman et al. has been incorporated, sometimes with minor differences, in different infectious settings.

In the field of urinary infections, although different studies have used adaptations of the definition of HA-UTI given by Friedman et al., to our knowledge not a single prospective study has evaluated the usefulness of these diagnostic criteria in UTI, and thus there is not an accepted definition for HA-UTI. The association of HA-UTI with QR suggests that the health care setting serves as a reservoir for QR and that patients acquire these QR E. coli resistant strains after establishing contact with it. This is supported by a study of Colodner et al. in which the authors observed a linear increase in the percentage of QR E. coli strains isolated from patients with CA-UTI depending on the number of previous hospitalizations [10]. Lautenbach et al. have recently demonstrated that gastrointestinal tract colonization with E. coli showing reduced susceptibility to levofloxacin is common among hospitalized patients, which also supports the concept that a previous hospitalization is a potential source of acquisition of a QR E. coli strain [22].

Another interesting observation of our study was that QR E. coli strains were more frequently resistant to other antimicrobials, compared to the QS E. coli strains, which has previously also been reported by other authors [6]. Resistance to quinolones in E. coli most commonly arises in a stepwise manner as a result of an accumulation of chromosomal mutations within the genes encoding the quinolone targets [23]. In Gram-negative bacteria, mutations first affect the gyrA gene and tend to be present at certain locations [24]. Additional chromosomal mutations that affect the accumulation of quinolones, including those affecting either the expression of porins or those involved in regulatory genes which affect the activity of a wide range of efflux pumps are also involved in the development of QR in E. coli isolates [23]. Mutations that lead to an increased expression of multidrug efflux pumps, particularly of the AcrAB, may explain the higher frequency of resistance to other antimicrobials found in QR E. coli strains [25]. In a recent study, 50% of the E. coli strains with reduced susceptibility to levofloxacin showed increased drug AcrAB efflux pump expression, higher than previously seen, which suggests that this mechanism of resistance may be becoming widespread over time [25]. Resistance to FQ can also be transmitted through plasmids [26]. Strains that posses the quinolone plasmids encoded resistance are frequently associated with CTX-M enzymes that inactivate third generation cephalosporins, which may partially explain the frequent association, as observed in our study, between QR and ESBL production in Enterobacteriaceae [26].

The other variable that was associated with QR in the multivariate analysis was having received antibiotic treatment, particularly FQ, in the preceding month. The impact of antibiotic prescription on antimicrobial resistance has been evaluated in a meta-analysis by Costelloe et al. in which the authors found that antibiotics prescribed in primary care were associated with subsequent UTI caused by E. coli QR strains, particularly when administered in the preceding month [27]. Prior antibiotic use with quinolones has widely been recognised as a factor related to QR in E. coli strains causing community UTI [8–10, 13]. It has been demonstrated that prolonged administration of FQ in patients with ABP leads to a substitution in the faeces of the QS E. coli strains, isolated before treatment, by genetically distinct QR E. coli strains [28]. It has been speculated that these QR E. coli strains are not the result of the selective pressure exerted by FQ on intermediate-susceptible E. coli strains but rather the consequence of unmasking QR E. coli present in low numbers in the gastrointestinal tract before treatment or the acquisition of an exogenous QR E. coli strain during treatment [28]. These QR E. coli strains, colonizing the gastrointestinal tract, could be the source of the infection in men with FUTI and recent intake of FQ, following a “fecal-urethral” pathway. The “fecal-urethral” pathway, which is the route of infection in women with UTI, is not clearly the way of acquisition of the infection in men according to a study of Johnson et al. in which only a minority of men with FUTI had the same E. coli strain in rectal isolates and in urine, suggesting that an alternative infection route or exacerbations of a pre-existing chronic bacterial prostatitis (CBP) could be involved in the pathogenesis of UTI in men [29]. In our study, nearly 40% of the patients referred a history of previous UTI, which rose to 60% in the group of FUTI caused by QR E. coli strains.

Recurrent UTI in men are frequently associated with the existence of a subjacent CBP [1]. Mazzoli has demonstrated that E. coli strains isolated from patients with CBP are frequently biofilm producers, which in turn, has been recognised as one of the main causes of recurrent infections [30]. Although the data regarding the antibiotic susceptibility profile is not shown in the article by Mazzoli it is suggested that biofilm producer’s strains showed high antimicrobial resistance, which could be related to a previous antimicrobial exposure. Patients with CBP usually receive long and recurrent antimicrobial courses, which frequently involve the administration of FQ. One might speculate that the administration of repeated therapeutic courses of FQ could expose E. coli strains present in the deep biofilm layers to sub-inhibitory concentrations of FQ, which could lead to the selection of QR strains that could cause subsequent episodes of UTI. In fact in our study a high proportion of patients with a FUTI caused by a QR E. coli strain had a previous history of UTI and had received FQ in the past month (data not shown).

The present study has some potential limitations that have to be taken into consideration. First of all, as previously mentioned, we have used a definition of HA-UTI that has not been supported by prospective studies. Second, we included as QR not only those E. coli strains resistant to ciprofloxacin but also those that were resistant to pipemidic acid and susceptible to ciprofloxacin while other studies, but not all [20], tend to evaluate only FQ E. coli resistant strains. Most E. coli isolates with diminished susceptibility to ciprofloxacin are resistant to nalidixic acid and carry a mutation in the gyrA gene [24]. E. coli strains susceptible to FQ that already carry a mutation in the gyrA gene can easily develop a second mechanism of resistance and therefore acquire higher levels of resistance in the presence of quinolones when compared to those not having the mutation [24]. The clinical importance of this eventuality has clearly been demonstrated in typhoid fever caused by nalidixic resistant and ciprofloxacin susceptible Salmonella typhi strains in which the treatment with FQ is associated with a high risk of therapeutic failures [31]. Based on this evidence it has been suggested that invasive infections caused by nalidixic resistant and quinolone susceptible Enterobacteriacea strains should not be treated with FQ particularly if there is a high bacterial inoculum or if these agents achieve low concentrations at the location of the infection [32]. The importance of treating with FQ urinary infections caused by E. coli strains resistant to nalidixic acid and susceptible to ciprofloxacin has not been properly evaluated although the fact that most quinolones, particularly levofloxacin, have a renal elimination, probably limits the importance of this effect. The third consideration is that we have analyzed an apparent miscellaneous group of patients with FUTI, as we included males with acute pyelonephritis, ABP and those with FUTI without topographic location of the infection. This approach could be questioned but as stated in the Guidelines on Urological Infections from the European Association of Urology, most men with FUTI have a concomitant infection of the prostate and therefore should be considered as an ABP [33].

The importance of this study is that it not only confirms the significance of previously known risk factors for QR in men, mainly the existence of a previous exposure to FQ, but also highlights the relevance of new risk factors such as having had a recent health contact. Although FQ are considered the optimal treatment in males with FUTI, risk factors for QR such as whether the patient has an HA-UTI or whether he has received antibiotic treatment with FQ in the preceding month should be taken into consideration before initiating empirical treatment with these agents, at least until antimicrobial susceptibility tests are available for the clinician.