Risk Factors for Community-Acquired Urinary Tract Infections Caused by ESBL-Producing Enterobacteriaceae –A Case–Control Study in a Low Prevalence Country

Community-acquired urinary tract infection (CA-UTI) is the most common infection caused by extended-spectrum β-lactamase (ESBL)-producing Enterobacteriaceae, but the clinical epidemiology of these infections in low prevalence countries is largely unknown. A population based case-control study was conducted to assess risk factors for CA-UTI caused by ESBL-producing E. coli or K. pneumoniae. The study was carried out in a source population in Eastern Norway, a country with a low prevalence of infections caused by ESBL-producing Enterobacteriaceae. The study population comprised 100 cases and 190 controls with CA-UTI caused by ESBL-producing and non-ESBL-producing E. coli or K. pneumoniae, respectively. The following independent risk factors of ESBL-positive UTIs were identified: Travel to Asia, The Middle East or Africa either during the past six weeks (Odds ratio (OR) = 21; 95% confidence interval (CI): 4.5–97) or during the past 6 weeks to 24 months (OR = 2.3; 95% CI: 1.1–4.4), recent use of fluoroquinolones (OR = 16; 95% CI: 3.2–80) and β-lactams (except mecillinam) (OR = 5.0; 95% CI: 2.1–12), diabetes mellitus (OR = 3.2; 95% CI: 1.0–11) and recreational freshwater swimming the past year (OR = 2.1; 95% CI: 1.0–4.0). Factors associated with decreased risk were increasing number of fish meals per week (OR = 0.68 per fish meal; 95% CI: 0.51–0.90) and age (OR = 0.89 per 5 year increase; 95% CI: 0.82–0.97). In conclusion, we have identified risk factors that elucidate mechanisms and routes for dissemination of ESBL-producing Enterobacteriaceae in a low prevalence country, which can be used to guide appropriate treatment of CA-UTI and targeted infection control measures.


Introduction
During the past 15 years, we have observed a worldwide dissemination of infections caused by CTX-M extended-spectrum b-lactamase (ESBL)-producing Enterobacteriaceae [1]. These infections are associated with increased mortality, morbidity, health care costs, and the need for broad-spectrum antibiotics [2]. Community-acquired urinary tract infection (CA-UTI) is the most common infection caused by ESBL-producing bacteria, but we have limited knowledge regarding the clinical epidemiology of these infections [1,3]. Most studies have focused on health care related infections and associated risk factors. Moreover, these studies have largely been based on information from medical records. Thus, information on possible risk factors not regularly noted in those records is sparse [4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20]. A large multinational survey of infections caused by ESBL-producing Enterobacteriaceae identified age $65 years, male sex and recent use of cephalosporins as independent risk factors for CA-ESBL infections [3]. However, the authors expressed a poor predictive value of their chosen model.
The present study was conducted in Norway. The yearly Norwegian nationwide antimicrobial resistance surveillance programme has shown a very low prevalence of infections caused by ESBL-producing Enterobacteriaceae [21]. A prevalence of 1.6% ESBL positive UTI in the Norwegian population was estimated for 2011. The prevalence is slowly increasing. A country with low prevalence of infections with ESBL-producing bacteria is well suited to identify risk factors for acquisition of ESBL, and a nationwide prescription database makes Norway suitable for the study of antibiotic use in detail [22]. Based on these advantages and patient interviews we aimed to investigate whether patients with ESBL positive CA-UTI have a different frequency of risk factors of CA-UTI as compared to patients with ESBL negative CA-UTI.

Design and Study Population
A case-control study was conducted at the Department of Medical Microbiology, Vestre Viken Hospital Trust situated in a mixed urban, suburban and rural area in the South-Eastern part of Norway. Our two laboratories analyse samples from in-and outpatients in an area comprising four hospitals and approximately 450.000 inhabitants (source population). The inclusion period was from February 2009 to April 2011.
The eligible population constituted all patients $18 years old with a urine culture yielding E. coli or K. pneumoniae .10,000 CFU/ ml. The following exclusion criteria were used: i) patients who had lived in Norway for ,1 year, ii) were unable to answer our questionnaire, iii) had previously diagnosed infection caused by ESBL-producing bacteria, and iv) patients with health care associated UTI (i.e., hospitalized or residing in a nursing home for .24 hours during the last 31 days).
The study population consisted of all patients willing to participate with ESBL-positive UTI (case group) and randomly selected patients with ESBL-negative UTI (control group) ( Figure 1).
The patients received written information and were invited to participate by ordinary mail. Non-responders were contacted twice. Acceptance was given by returning a signed consent form.
Ethics statement. The study was approved by the Regional Committee for Medical and Health Research Ethics in South-Eastern Norway (reference number: 2009/2037 BS-08901b).

Data Collection
Urine cultivation and bacterial identification were performed using ChromID CPS3 agar and the VITEK-2 system (both BioMerieux, Marcy l'Etoile, France). Antimicrobial susceptibility testing and interpretations including ESBL screening were performed using VITEK-2 or agar disc diffusion method according to EUCAST recommendations and clinical breakpoints [23].
Isolates resistant to cefpodoxime, cefotaxime or ceftazidime were selected for confirmatory ESBL testing using the E-test system (AB-Biodisk, BioMerieux). ESBL genotype analysis was performed using PCR for bla CTX-M detection and group assignment, as described [24]. Isolates negative for bla CTX-M were analyzed using conventional bla TEM and bla SHV PCR and sequencing, as described [25].
A structured interview was performed by a trained investigator by telephone or in-person for community-based and hospitalized patients, respectively. The questionnaire was sent to the participants in advance and included questions regarding the infection for which they were included in the study, health condition (Charlson Comorbidity Index [26]), contact with the health care system in Norway and abroad (time and duration during the past 5 years), UTIs, antibiotic use, compliance with antibiotic prescriptions, prostate disease, use of a urinary catheter during the past year, oral and digestive health problems, international travel or residency lasting $24 hours during the past five years (time since returning home, duration and country), profession, personal hygiene, household members, pets, eating habits (meals per week of different foods and meals outside home), and recreational swimming during the past year (location, number of times and submergence of head).
In Norway antibiotics are available on prescription only. Date, type and amount of antibiotic dispensed during the past five years were obtained from The Norwegian Prescription Database [22]. Information about antibiotic use during hospitalization was obtained from medical records.
Information on previous infections with ESBL-producing bacteria was obtained from our laboratorỳs computer system.

Statistical Analysis
This case-control study was analysed using a pragmatic strategy, which means that priority was not given to a specific hypothesis.
Univariate analyses were performed using Student's t test, Pearson's chi-square test or Fisher's exact test when appropriate. The association between potential risk factors and infection caused by ESBL-producing E. coli or K. pneumoniae was quantified by odds ratio (OR) with 95% confidence interval (CI). Any variable with a p,0.15 from the univariate analysis was considered a candidate for the multivariate model. A manual backward stepwise Figure 1. Selection of study population. a Dementia (n = 1), unable to reach by phone (n = 2) and death (n = 2). doi:10.1371/journal.pone.0069581.g001 elimination procedure using a multivariate logistic regression model was performed to identify independent risk factors. Multivariate analyses were preceded by estimation of correlation between risk factors. Evaluation of the predictive accuracy of the models was assessed by calibration and discrimination. Calibration was evaluated by the Hosmer and Lemeshow goodness-of-fit test. A statistically non-significant Hosmer and Lemeshow result (p.0.05) suggests that the model predicts accurately on average. Discrimination was evaluated by analysis of the area under the ROC curve. We defined acceptable discriminatory capability as an area under the ROC curve greater than 0.7 [27]. Two-tailed p values of ,0.05 were considered statistically significant. All statistical analyses were conducted using PASW statistics software, version 19.0 (IBM SPSS, Chicago, IL).

Results
Approximately 28,000 urine samples from 15,000 unique patients were submitted to our department during the inclusion period. A total of 359 (1.3%) samples yielded ESBL positive E. coli (n = 342) or K. pneumoniae (n = 17). After exclusion 171 subjects with ESBL UTI were invited to participate (case group). Also, 439 randomly selected control patients were invited to participate ( Figure 1).
Relevant background characteristics of the participants are presented in Table 1. The cases and controls were in large similar.
Significantly younger age and the presence of diabetes mellitus among cases were the two exceptions.

ESBL Genotyping
PCR and sequence analyses showed that 65%, 30%, and 5% of the ESBL isolates belonged to the CTX-M group 1, CTX-M group 9 and SHV group 5/12, respectively. TEM-type ESBLs were not detected.

Antibiotic Use and Antibiotic Resistance
Data on antibiotic use are presented in Table 2. More than 90% of the participants reported that they had completed all prescribed courses of antibiotics received during the past 5 years. Antibiotic use was more prevalent in the study population (59% during the past three months before the infection) than in the age-adjusted general Norwegian population (29% during the past year) -(data from the Norwegian Prescription Registry [22]). This difference was mainly due to increased use of antimicrobials used to treat UTIs in the study population.
In general, ESBL-producing isolates expressed more coresistances compared to non-ESBL strains. For cases and controls the proportion of non-susceptible strains were 59% and 13% for ciprofloxacin, 78% and 24% for trimethoprim, 35% and 4% for gentamicin, 4% and 2% for nitrofurantoin and 4% and 3% for mecillinam, respectively.

Risk Factor Analysis
The results of the univariate analyses on risk factors are presented in Table 3. Travelling to Asia, Middle East or Africa up to 2 years in the past, recreational swimming, eating dinner at restaurants and close occupational contact with humans were identified as significant risk factors for ESBL UTI. Interestingly, frequent consumption of fish meals (Figure 2), infrequent bath or shower and digestive problems seemed to have a protective effect.
The results of the multivariate analyses are presented in Table 4. Patients with an ESBL positive UTI had travelled 21 times more to Asia, Middle East or Africa during the past 6 weeks than patients with a non-ESBL UTI, and this was the strongest predictor for ESBL UTI. Travel to the same areas in the period from 6 weeks to 24 months in the past was to a lesser degree associated with ESBL UTI (OR 2.3, 95% CI: 1.2-4.4, p = 0.017). The variables regarding (time since) travel abroad were also analysed as continuous variables but this did not influence the results. Recreational freshwater swimming was identified as an independent risk factor, and patients with ESBL UTI had swum twice as frequent in freshwater as patients with ESBL negative UTI.
Previously known risk factors such as recent antibiotic use and diabetes mellitus were also identified as independent risk factors. Age and weekly fish meals were found to be putative protective factors.
The final multivariate model was applied to participants with infection caused by E. coli only and this did not change any trends in the results (data not shown).
The Hosmer and Lemeshow goodness-of-fit test was not significant indicating a satisfactory fit of the model (x 2 = 5.64, df = 8, p = 0.69). The area under the ROC curve was 0.83 (95% CI: 0.79-0.88) indicating a good discriminative ability between ESBL-positive and ESBL-negative patients.

Discussion
This is to our knowledge the first population-based study to identify risk factors for acquisition of CA-ESBL infections in a low prevalence country. International travel was identified as the most important risk factor for ESBL positive CA-UTI in this study. Most travel-associated ESBL UTIs occurred during the first six weeks after returning home. This observation is consistent with previous studies and adds new information about the time course between colonization during travel and actual infection [5,28,29]. The area associated with the highest risk (Asia, Middle East and Africa) corresponds well with areas previously associated with a high rate of colonization in returning travellers [28].
This observation contrasts a recent French study. Nicolas-Chanoine and co-workers did not identify travelling abroad for  Data are presented as the absolute number of patients with percentages in parentheses with the exception of fish and meat meals, which is listed as mean value 6 SD. b Some variables have missing values (number of missing patients in parentheses). Usually submerges head during recreational swimming (7), Organic food (7), Dinner in restaurant (2), Prefers meat well done (8), Close occupational contact with humans (6), Bath or shower (3), Digestive problems (6)  .14 days during the past 6 months as a risk factor for an ESBLpositive (bla CTX-M-15 ) infection in hospitalized patients [6]. In our study, travelling abroad for .14 days was a strong predictor of ESBL UTI when using bla CTX-M-1 -positive infections as an endpoint (data not shown). It is likely that the importance of travel as a risk factor will differ between the French hospitalized population and the Norwegian non-hospitalized population in our study. Also, the proportion of ESBL-producing clinical isolates of Enterobacteriaceae in France is higher than in Norway [30]. Therefore, travel abroad from France will not have the same relative impact on the colonization and infection rate as travelling abroad from Norway. This emphasizes the importance of investigating these risk factors in a low prevalence area. Recent antibiotic use is a known risk factor for infections caused by ESBL-positive bacteria [3,7,8,11,31]. We found that recent use of fluoroquinolones was strongly associated with an ESBL-positive UTI, supporting the results from several other studies [7,8,31].
Interestingly, the use of mecillinam as opposed to other blactams, was not associated with ESBL-positive CA-UTI. This may be because the oral formulation of mecillinam, pivmecillinam, is a pro-drug with minor effects on the intestinal flora [32]. Moreover, mecillinam has a selective activity against Gramnegative bacteria and is more stable against ESBL hydrolysis compared to most penicillins [33].
Recreational swimming in freshwater was identified as an independent risk factor for ESBL UTI. ESBL-producing bacteria like E. coli have been detected in environmental water [34][35][36]. Furthermore, outbreaks of E. coli O157:H7 have been linked to swimming in contaminated freshwater [37]. Swimming may therefore be a risk factor for intestinal colonization with E. coli with ESBL and any subsequent UTI may be caused by a such newly acquired ESBL-producing strain from the water [38]. This finding highlights a possible link between environmental pollution and antimicrobial resistance, but will have to be substantiated before any conclusions can be drawn [39].
Interestingly, eating fish was associated with a reduced risk of ESBL UTI (Figure 2). Each weekly fish meal reduced the risk of an ESBL positive infection with about 30%. It is clear that eating habits influence the microbial flora in the gut [40]. However, whether eating fish may affect the resistance pattern of the gut microbial flora and potentially lower the risk of ESBL UTI remains speculative and eating fish may be a marker of a more fundamental risk factor not measured.
Retail chicken meat has recently been implicated as a possible source of ESBL-colonization [41]. We did not specifically investigate this possible risk factor, but ESBL-producing bacteria have only very rarely been found in the Norwegian food chain [42].
In our study, patients infected with an ESBL-producing E. coli or K. pneumoniae were significantly younger than the control patients. In two studies with similar design but including hospitalized patients, no association between age and ESBL positive infection was found [8,43]. This suggests that the epidemiology of ESBL infections differs in Norway or among non-hospitalized patients.

Limitations
Limitations include the possibility of selection bias due to nonparticipation and a potential problem with differential misclassification of exposure because the interviewers were not masked to the status of the patient being a case or a control. To minimize the latter the questionnaires were sent to the participants in advance and interview training was given.
We did not use the Friedman criteria for health care acquired infections and thus patients with health care system contact during the past 2-3 months and patients catheterized the past month were included for analysis [44]. Excluding these patients (n = 30) did, however, not change any trends in the results (data not shown).
Finally, our study may overestimate the use of antibiotics as a risk factor since patients in the control group, with susceptible bacteria, may be less likely to have used antibiotics. This is because non-ESBL E. coli and K. pneumoniae are more susceptible to antibiotics than ESBL-producers, and recently treated patients with such susceptible strains are therefore less likely to show up in the control group [45].
In summary, we have addressed the knowledge gap concerning risk factors for CA-UTIs caused by ESBL-producing bacteria [3]. Previously suspected risk factors for ESBL UTI have been supported and possible new ones uncovered. Our study shows that the predictive antimicrobial resistance pattern in uropathogenic E. coli is heavily influenced by the country the patient has recently visited [28,46]. Thus, information on recent travel is important when treating patients with serious infections that may involve this organism. Physicians in low-prevalence countries should consider ESBL when treating UTI in patients who have visited countries in Africa, The Middle East or Asia during the past six weeks [28,46].
An association between recreational swimming and ESBL UTI was detected. Further investigation to examine the possible negative impact of environmental pollution with ESBL-producing Enterobacteriaceae seems warranted.
Finally, eating fish regularly was associated with a protective effect against ESBL UTI. If this is confirmed in other studies, an interesting link between diet and infection has been established.