We determined host and pathogen risk factors for urinary-source bacteremia in a prospective study of patients with Escherichia coli bacteriuria. Both host (urinary retention; history of urogenital surgery) and pathogen factors (a capsule characteristic) were independent predictors of bacteremia.
Abstract
Background. The urinary tract is the most common source for Escherichia coli bacteremia. Mortality from E. coli urinary-source bacteremia is higher than that from urinary tract infection. Predisposing factors for urinary-source E. coli bacteremia are poorly characterized.
Methods. In order to identify urinary-source bacteremia risk factors, we conducted a 12-month prospective cohort study of adult inpatients with E. coli bacteriuria that were tested for bacteremia within ±1 day of the bacteriuria. Patients with bacteremia were compared with those without bacteremia. Bacterial isolates from urine were screened for 16 putative virulence genes using high-throughput dot-blot hybridization.
Results. Twenty-four of 156 subjects (15%) had E. coli bacteremia. Bacteremic patients were more likely to have benign prostatic hyperplasia (56% vs 19%; P = .04), a history of urogenital surgery (63% vs 28%; P = .001), and presentation with hesitancy/retention (21% vs 4%; P = .002), fever (63% vs 38%; P = .02), and pyelonephritis (67% vs 41%; P = .02). The genes kpsMT (group II capsule) (17 [71%] vs 62 [47%]; P = .03) and prf (P-fimbriae family) (13 [54%] vs 40 [30%]; P = .02) were more frequent in the urinary strains from bacteremic patients. Symptoms of hesitancy/retention (odds ratio [OR], 7.8; 95% confidence interval [CI], 1.6–37), history of a urogenital procedure (OR, 5.4; 95% CI, 2–14.7), and presence of kpsMT (OR, 2.9; 95% CI, 1–8.2) independently predicted bacteremia.
Conclusions. Bacteremia secondary to E. coli bacteriuria was frequent (15%) in those tested for it. Urinary stasis, surgical disruption of urogenital tissues, and a bacterial capsule characteristic contribute to systemic invasion by uropathogenic E. coli.
The urinary tract is the most common source of systemic Escherichia coli infection. Escherichia coli bacteremia is associated with a mortality rate of 5%–21% [1, 2]. Antimicrobial-resistant isolates can result in mortality rates as high as 61% [3]. Approximately 36 000 deaths occur in the United States annually from E. coli bacteremia [4].
Multiple studies have attempted to determine risk factors for the progression from E. coli urinary tract infection (UTI) to urinary-source bacteremia (USB) [5–11]. However, most were limited to community-acquired UTIs [5–7, 11] or female patients [5, 7] or were conducted retrospectively [8, 10]. In addition, not all studies described if they used UTI patients with negative blood cultures for comparison [9, 10]. Only 1 study had >100 patients: Velasco et al conducted a prospective study of 669 community-acquired infections [11]. The only consistent, clinical risk factor that emerged from these studies was advanced patient age [5–7, 10, 11]. However, clinical factors are not the only determinant of risk of bacteremia; the inherent virulence of the microorganism may also play a role. Previous studies have reported inconsistent associations with specific bacterial virulence factors and were limited by small sample sizes [5–10], with a maximum of 100 subjects [8].
Our study objective was to identify host and pathogen risk factors for E. coli USB in a large, inclusive population. Our long-term goal is to contribute to the clinical decision making for hospital patients with UTIs and to help improve outcomes. Identification of novel risk factors could contribute to predictive models that facilitate early recognition of high-risk patients.
METHODS
Study Design, Data Collection, and Definitions
We conducted a prospective cohort study of patients with E. coli bacteriuria from 1 August 2009 until 31 July 2010 at Barnes-Jewish Hospital, a 1250-bed teaching hospital in Missouri. All adult patients admitted to Barnes-Jewish Hospital who presented with or developed E. coli bacteriuria during their hospital stay and had blood cultures taken at time of bacteriuria were eligible for enrollment. Patients with polymicrobial UTIs and/or concurrent bloodstream infection with an organism other than E. coli were excluded. Medical records of those who met inclusion criteria were reviewed for demographics and medical/urogenital history. Charlson comorbidity and McCabe severity-of-illness scores were computed. The patients’ clinical presentation; vital signs; and laboratory, radiological, and pharmacy data were prospectively reviewed during the admission. For each antibiotic with gram-negative activity, the start and stop times were recorded.
The primary outcome was development of E. coli bacteremia. Blood cultures were drawn at the discretion of the treating physicians and had to occur within ± 1 day of the positive urine culture. Given that urine and blood specimens were processed almost simultaneously, the primary outcome was determined within 1 day of enrollment. Secondary outcomes were sepsis, sepsis-induced hypotension, transfer to the intensive care unit (ICU) within 3 days, length of hospital stay after detection of bacteriuria, and in-hospital mortality. At the time of the study, the cutoff for significant bacteriuria used by the hospital microbiology laboratory was 5 × 104 colony-forming units (CFU)/mL in noncatheterized patients and 5 × 103 CFU/mL in catheterized patients. Bacteriuria was classified as community-acquired if the first positive urine culture occurred ≤48 hours after admission.
Bacteriuria was classified using the patients’ documented urinary symptoms. Asymptomatic bacteriuria was defined as absence of urinary symptoms and/or fever; cystitis was defined as presence of dysuria, frequency, or urinary retention/hesitancy (without signs of pyelonephritis) [12]; pyelonephritis was defined as presence of flank pain or tenderness and/or fever; and unclassified bacteriuria was defined as bacteriuria in a patient who did not fit any of the above criteria and could not report symptoms (eg, intubation, altered mental status). The bacteriuria was considered to be catheter-associated if a urinary catheter had been in place in the 48 hours prior to the positive urine culture; the same classification criteria as above were used. Past urogenital surgery included all surgeries that resulted in anatomical alteration (eg, nephrectomy, neobladder formation, hysterectomy, prostate resection). Sepsis and sepsis-induced hypotension were defined using established criteria [13]. Adequacy of antibiotic therapy was defined as pathogen-directed treatment with matching antibiotic susceptibilities.
Laboratory Analyses
We identified the E. coli isolates from urine cultures of eligible patients in the hospital microbiology laboratory and stored them at −80°C in skim milk. The isolate collection was shipped to the Molecular and Clinical Epidemiology Laboratory at the School of Public Health, University of Michigan, Ann Arbor, for further processing. Bacterial DNA was extracted using QIAamp DNA mini kit (Qiagen, Valencia, California). DNA probes for virulence genes were designed as previously described [14–18] (Table 1). The presence of these virulence genes was determined by dot-blot hybridization with fluorescent-labeled probes and a fluorescein-based detection system as described elsewhere [19]. To screen large numbers of isolates in short time periods, we used a microarray system for high-throughput hybridization [20, 21]. The frequency of specific virulence genes was compared between the urine isolates of bacteremic and nonbacteremic patients (ie, bacteriuria with negative blood cultures). Bacterial isolates were compared regarding antimicrobial susceptibility patterns obtained using disk diffusion tests.
Table 1.
Prevalence of Virulence Genes Tested in This Study
| Gene Name (in Order of Frequency Among Bacteremic Isolates) | Abbreviation | function | Total (N = 156), No. (%) | Bacteremic Isolates (n = 24), No. (%) | Nonbacteremic Isolates (n = 132), No. (%) | P Value |
|---|---|---|---|---|---|---|
| E. coli heme utilization | chuA | Heme uptake | 132 (85) | 23 (96) | 109 (83) | .1 |
| Outer membrane protein | ompT | Surface protease | 132 (85) | 22 (92) | 110 (83) | .4 |
| Ferric yersiniabactin uptake | fyuA | Siderophore (yersiniabactin) | 130 (83) | 21 (88) | 109 (83) | .8 |
| Hypothetical protein | yjaH | Unknown | 122 (78) | 17 (71) | 105 (80) | .3 |
| Anonymous DNA fragment | tspE | Unknown | 115 (74) | 17 (71) | 98 (74) | .7 |
| Uropathogenic specific protein | usp | Bacteriocin | 110 (71) | 17 (71) | 93 (71) | >.99 |
| Group II capsule characteristic | kpsMT | Capsule antigen | 79 (51) | 17 (71) | 62 (47) | .03 |
| P-related fimbriae | prf | Adhesin | 53 (34) | 13 (54) | 40 (30) | .02 |
| Iron uptake chelate | iucD | Siderophore (aerobactin) | 71 (46) | 11 (46) | 60 (46) | >.99 |
| IrgA homologue adhesin | iha | Adhesin | 62 (40) | 10 (42) | 52 (39) | .8 |
| Secreted autotransporter toxin | sat | Toxin | 60 (39) | 9 (38) | 51 (39) | .9 |
| Hemolysin | hlyA | Toxin | 41 (26) | 6 (25) | 35 (27) | .9 |
| S-fimbrial adhesin | sfa | Adhesin | 37 (24) | 5 (21) | 32 (24) | .7 |
| Cytotoxic necrotizing factor | cnf1 | Toxin | 27 (17) | 5 (21) | 22 (17) | .6 |
| Salmochelin | iroN | Siderophore | 39 (25) | 4 (17) | 35 (27) | .3 |
| Dr family of adhesins | Dr | Adhesin | 10 (6) | 10 (8) | 0 | .4 |
Sample Size Calculations and Statistical Analysis
We assumed that there would be 15% bacteremic vs 85% nonbacteremic cases [5]. Based on data regarding the papGiA2 genotype (a virulence gene associated with E. coli bacteremia in an earlier study) [6], we estimated that bacteremic E. coli will carry prf (P-related fimbriae, a previously described marker for a variety of P-fimbriae [22]) in 70% of cases and nonbacteremic E. coli in 35%. To determine a difference at the .05 significance level with 80% power, we estimated that we would need at least 21 bacteremic vs 119 nonbacteremic isolates (EpiInfo, version 3.3.2).
Data analysis was performed using SPSS 18 (SPSS, an IBM Company, Chicago, Illinois). Univariate comparisons among categorical variables were performed using the χ2 test or Fisher exact test as appropriate. Comparisons among continuous independent variables were performed using Student t test or Mann-Whitney U test as appropriate. A 2-sided P value of <.05 was considered significant. Variables with a P value of <.1 on univariate testing were entered into a multivariate logistic regression model in 1 step.
The study was approved both by the Washington University Human Research Protection Office and the University of Michigan Institutional Review Board.
RESULTS
We identified 337 patients with E. coli bacteriuria during the study period. Of these, 181 (54%) were excluded due to absence of blood cultures in the predetermined time window, leaving 156 patients with bacteriuria who had blood cultures drawn (Table 2). Study patients were primarily female (n = 111; 71%) and white (n = 91; 58%); the median age was 66 (range, 19–98) years. Twenty-one (14%) patients were admitted from a long-term care facility. Most patients (n = 103; 66%) had community-acquired bacteriuria. Forty-two (27%) had asymptomatic bacteriuria; 14 (9%) had cystitis; 70 (45%) had pyelonephritis (with or without cystitis); and 30 (19%) were labeled as having unclassified bacteriuria. Thirty-three (21%) patients had urinary catheter–associated bacteriuria (4 with cystitis, 11 with pyelonephritis, 9 with asymptomatic bacteriuria, and 9 with unclassified bacteriuria).
Table 2.
Comparison of 156 Bacteriuric Patients With or Without Escherichia coli Bacteremia
| Variable | Bacteremia (n = 24) | No Bacteremia (n = 132) | P Value | Adjusted OR (95% CI)a |
|---|---|---|---|---|
| Sex (female) | 15 (63) | 96 (73) | .3 | |
| Age in years, median (range) | 67.5 (23–92) | 65 (19–98) | .4 | |
| Race (white) | 17 (71) | 74 (56) | .2 | |
| Body mass index, kg/m2 (SD) | 27.0 (5.7) | 28.5 (8.6) | .4 | |
| Diabetes mellitus | 8 (38) | 48 (36) | .9 | |
| Renal insufficiency (Cr > 1.5 mg/dL) | 9 (38) | 31 (24) | .1 | |
| Any malignancy | 6 (25) | 36 (27) | .8 | |
| Any transplant | 2 (8) | 8 (6) | .7 | |
| Benign prostatic hyperplasia (male patients) | 5/9 (56) | 7/36 (19) | .04 | |
| History of urogenital surgery | 15 (63) | 37 (28) | .001 | 5.4 (2.0–14.7) |
| Urological procedure this admission | 1 (4) | 2 (2) | .4 | |
| Charlson comorbidity index, median (range) | 3 (0–9) | 3 (0–12) | .5 | |
| McCabe severity-of-illness score, median (range) | 1 (1–2) | 1 (1–3) | .4 | |
| Dysuria | 6 (25) | 16 (12) | .1 | |
| Frequency/urgency | 4 (17) | 20 (15) | .8 | |
| Hesitancy/retention | 5 (21) | 5 (4) | .002 | 7.8 (1.6–37.0) |
| Fever | 15 (63) | 50 (38) | .02 | |
| Confusion; altered mental status | 12 (50) | 46 (35) | .2 | |
| Sepsis | 20 (83) | 92 (70) | .2 | |
| Sepsis-induced hypotension | 14 (58) | 36 (27) | .003 | |
| Asymptomatic bacteriuria | 3 (13) | 39 (30) | .08 | |
| Cystitis | 2 (8) | 12 (9) | >.99 | |
| Pyelonephritis | 16 (67) | 54 (41) | .02 | |
| Unclassified bacteriuria | 3 (13) | 27 (21) | .4 | |
| Community-acquired bacteriuria | 18 (75) | 85 (64) | .3 | |
| Urinary catheter–associated bacteriuria | 2 (8) | 31 (24) | .1 | |
| Urinalysis with pyuria (>10 WBC) | 19 (91) | 84 (71) | .2 | |
| prf (P-fimbriae family) | 13 (54) | 40 (30) | .02 | 2.6 (.98–7.1) |
| kpsMT (group II capsule) | 17 (71) | 62 (47) | .03 | 2.9 (1.0–8.2) |
| Length of hospital stay in days, median (range) | 6.0 (3–40) | 5.0 (0–54) | .1 | |
| In-hospital mortality | 2 (8) | 12 (9) | >.99 |
Data are No. (%) unless otherwise specified.
Abbreviations: CI, confidence interval; Cr, creatinine; OR, odds ratio; SD, standard deviation; WBC, white blood cell.
Variables considered for inclusion in the final model included history of urogenital surgery, hesitancy/retention, prf genotype, and kpsMT genotype.
The median Charlson comorbidity index was 3 (range, 0–12) with the most frequent comorbidities being diabetes mellitus (37%), malignancy (27%), and renal insufficiency (26%). Among 111 female patients, 3 (3%) were peripartum. In 45 male patients, 12 (27%) had a diagnosis of benign prostatic hyperplasia, and 5 (11%) had a history of prostatic carcinoma. Sixty-nine (44%) subjects had records indicating ≥1 previous UTI. The median McCabe severity-of-illness score was 1 (range, 1–3). Fourteen (9%) patients died during hospital admission.
Patient Risk Factors for USB
Twenty-four of 156 subjects (15%) had E. coli bacteremia. Bacteremic patients were more likely than nonbacteremic patients to have a history of benign prostatic hyperplasia (5 of 9 [56%] vs 7 of 36 [19%]; P = .04) (of 45 male patients), prior urogenital surgery (15 of 24 [63%] vs 37 of 132 [28%]; P = .001), symptoms of hesitancy/retention (5 [21%] vs 5 [4%]; P = .002), and fever (15 [63%] vs 50 [38%]; P = .02) and were diagnosed more frequently with pyelonephritis (16 [67%] vs 54 [41%]; P = .02) (Table 2). Hesitancy/retention and benign prostatic hyperplasia were correlated (P = .049). Bacteremic and nonbacteremic patients had similarly low rates of urological procedures performed earlier during the current admission (1 [4%] vs 2 [2%]; P = .4). Patients were similar in terms of comorbidities and severity of illness as measured by the Charlson index and McCabe score.
In a stratified analysis by sex, hesitancy/retention was not associated with bacteremia in women when numbers were small (1 of 15 [7%] bacteremic vs 2 of 96 [2%] nonbacteremic patients; P = .3), but was associated in men (4 of 9 bacteremic [44%] vs 3 of 36 nonbacteremic patients [8%]; P = .02). Age was not associated with bacteremia in either sex (data not shown).
Bacterial Virulence Factors in USB
Overall, the most frequently detected virulence genes were ompT (132; 85%), chuA (132; 85%), fyuA (130; 83%), and yjaH (122; 78%). The gene kpsMT (group II capsule) was found in 79 (51%) and prf was found in 53 (34%) patients. Bacteremic isolates were more likely to carry kpsMT (17 [71%] vs 62 [47%]; P = .03) and prf (13 [54%] vs 40 [30%]; P = .02] genes than other virulence genes.
In addition, we compared virulence factor profiles (individual sequences of detected virulence factors) and found a large variety of profiles (>100). Only 1 profile (signature: tspE, fyuA, usp, yjaH, kpsMT, ompT, prf) was associated with bacteremia (3 [13%] vs 2 [2%]; P = .03], but it was present in only 5 patient isolates. The median aggregated number of virulence factors (out of 16) was 8.5 (range, 1–14) for bacteremic and 9 (range, 1–15) for nonbacteremic strains (P = .7).
When comparing resistance patterns, there was no difference in phenotypical ciprofloxacin resistance in bacteremic E. coli (4 [17%] vs 47 [36%]; P = .07].
Comparison Among Patients With Pyelonephritis or Symptomatic Bacteriuria
Because pyelonephritis is classically associated with USB, we conducted a subanalysis of these 70 cases. Sixteen (23%) pyelonephritis patients were bacteremic. Among patient-level risk factors, only a history of past urogenital surgery was more frequent in bacteremic patients (11 of 16 [69%] vs 17 of 54 [32%]; P = .008). Age did not differ significantly (data not shown). Among virulence factors, prf (9 of 16 [56%] vs 13 of 54 [24%]; P = .02) and chuA (16 of 16 [100%] vs 41 of 54 [76%]; P = .03) were present in more bacteremic than nonbacteremic cases. Mortality did not differ (0 of 16 [0%] vs 3 of 54 [6%]; P = >.99).
Increasing Prevalence of kpsMT but Not prf According to Infection Severity
The prevalence of kpsMT increased in isolates from patients with more invasive infections. In isolates from cystitis cases, kpsMT was 29% prevalent (4 of 14); in pyelonephritis, it was 54% prevalent (38 of 70); and in USB, it was 71% prevalent (17 of 24) (χ2 test for trend, P = .04). For prf, the corresponding prevalences were 43% (6 of 14) for cystitis, 31% (22 of 70) for pyelonephritis, and 54% (13 of 24) for USB (P = .1). The prf gene was less prevalent in pyelonephritis than in USB (P = .047). The averaged cumulative number of virulence factors did not increase depending on clinical manifestation (data not shown).
Multivariate Analysis of Risk Factors for USB
We included 4 variables in the final regression model. Symptoms of hesitancy/retention (odds ratio [OR], 7.8; 95% confidence interval [CI], 1.6–37), history of a urogenital procedure (OR, 5.4; 95% CI, 2–14.7), and presence of kpsMT (OR, 2.9; 95% CI, 1–8.2) independently predicted bacteremia. The geneprf did not remain a significant predictor (OR, 2.6; 95% CI, .98–7.1). The Hosmer-Lemeshow test indicated a good fit for the data (P = .3). The c statistic was 0.8.
Clinical Outcomes
The length of hospital stay postbacteriuria was similar between bacteremic and nonbacteremic patients (median, 6 days; range, 3–40 vs median, 5 days; range, 0–54; P = .1) as was in-hospital crude mortality (2 [8%] vs 12 [9%]; P = >.99). Bacteremic patients were, however, more likely to develop sepsis-induced hypotension (14 [58%] vs 36 [27%]; P = .003) and were more frequently transferred to an ICU (13 [54%] vs 23 [17%]; P < .001). We did not examine the effect of antibiotic selection as part of this analysis; however, the time from bacteriuria to appropriate antibiotics was shorter in bacteremic (median, 2 hours vs 5 hours; P = .02) than in nonbacteremic patients.
DISCUSSION
Escherichia coli bacteriurias are a first step toward kidney infection and USB. Here, we report findings from a prospective cohort of patients with E. coli bacteriuria at a tertiary care hospital, where we identified novel patient characteristics and bacterial virulence traits predictive for concurrent bacteriuria and USB. The strongest clinical predictors of bacteremia were hesitancy/retention symptoms in men and sex-independent urogenital surgery history. Group II capsule (kpsMT) was found more frequently among urinary isolates from individuals with bacteremia.
Previously, the only patient characteristic associated with USB was older age [6, 10]. Age was not associated with USB in this cohort, but symptoms of hesitancy and/or urinary retention at the time of bacteriuria were independent predictors among men. This has not been reported previously; but urinary retention is a known risk factor for developing a UTI in men [23, 24]. It is possible that longer exposure to bacteria not only increases the risk of local infection but also the risk of urothelial invasion. A history of benign prostatic hyperplasia also was associated with USB among men. Benign prostatic hyperplasia is thought to predispose to UTIs [25]. Further, the bacterial virulence factors required to colonize a structurally altered prostate may increase the risk of febrile UTIs [26].
Among men and women in this cohort, a past history of urogenital surgery was associated with USB. We found no other reports of this association in the literature. Iatrogenic modification of the urogenital tract might lead to impaired urinary passage or voiding and increased infection risk. Conversely, patients with preexisting voiding difficulties may have been more likely to require surgery in the first place. Ikaheimo et al reported that bacteremic patients were more likely to have a “compromising condition,” which by their definition included urological abnormalities [7]. Of note, we found no association with bacteremia in patients who had developed bacteriuria after a urogenital procedure during the same admission.
Several relatively small studies (≤100 isolates) have attempted to identify virulence factors that predict bacteremic progression by uropathogenic E. coli; almost all found P-fimbriae subtypes to play a role [5, 10, 27]. Other virulence determinants were not consistently associated with USB [8, 27]. Although prf was associated with an almost 2-fold increase in USB in univariate analysis of our cohort, it did not remain statistically significant after adjustment for hesitancy/retention and history of urogenital surgery. In contrast to these previous studies, ours is the first to report an association of kpsMT with USB risk in adults. A previous study among adults found no association with kpsMT [8], but a study among male infants identified the K1 antigen (detected by antiserum tests) either alone or together with the hly genotype as predictive for USB [28]. The kpsMT genes direct polysialic acid capsule antigen K1 synthesis [29], which was first identified in invasive E. coli strains [30]. K1 has been proposed to allow intracellular E. coli to evade lysis by preventing lysosomal fusion [31]. The kpsMT gene does not, however, detect K1 specifically but rather all group II K antigens; expression studies targeting K1 could be crucial in the next analytical step.
Increasing gram-negative antibiotic resistance has renewed interest in antivirulence therapeutic approaches [32]. In E. coli, K1 has been considered as a target for antivirulence compounds [33]. Assuming a higher prevalence of capsule antigen K1 expression in uropathogens with high bacteremic potential, a vaccine directed against capsule or epitopes extending beyond the capsule surface may be protective against USB. Indeed, cell surface vaccine candidates masked by capsule may be ineffective at preventing bacteremia [34].
Another interesting finding is the high prevalence of the Yersinia high-pathogenicity island (ie, fyuA) in 88% of USB isolates. FyuA has been shown to be more frequent in symptomatic urinary isolates (89%–94%) than in fecal, colonizing strains (30%–47%) and has emerged from Bayesian network analysis as a close correlate to the clinical presentation [35, 36]. These findings suggest this high-pathogenicity island as another reasonable target for a virulence-directed therapeutic.
Some authors have demonstrated that USB is a useful indicator of increased disease severity and postulated that its identification can help shape management [37]. Other studies have questioned the need for blood cultures and the benefit of diagnosing bacteremia in women with pyelonephritis [38, 39], presumably because outcomes were similar. Although we saw no mortality difference between bacteremic and nonbacteremic patients in our study, there were differences in management (eg, greater likelihood to be transferred to the ICU and shorter time to appropriate antibiotics for bacteremic patients) that may account for not seeing a difference in outcomes in this inpatient population. This goes also for variability in antibiotic choice, which we did not correct for in our outcomes analysis. Our findings differ from previous studies that only observed deaths in bacteremic patients [37, 40].
There are a number of limitations to our study. First, it was conducted at a tertiary care center, and the findings may not be generalizable to other settings. Because inclusion required a physician-ordered blood culture, septic or febrile bacteriuric patients at risk for bacteremia may have been missed. More than half of the potentially eligible population was excluded because no blood cultures were ordered. Because greater severity of illness may have prompted a blood culture, it is possible that our nonbacteremic comparator group exhibits more similarities to the USB group than if all bacteriuric patients had been included. This may have diminished our ability to detect associations between patient characteristics, E. coli virulence factors, and USB. Symptoms at the time of bacteriuria were taken from hospital charts entered by physicians and nurses; patient symptoms may have been missed or not recorded. If recording of symptoms was associated with USB, it would bias our associations away from the null hypothesis. Lastly, although we started with the assumption that E. coli bacteremia is secondary to a urinary source, the close temporal relationship between isolates could also be consistent with “descending” infection of the urinary tract from the bloodstream.
In summary, our study is the largest cohort study to analyze both patient and pathogen factors associated with USB and the first to be conducted prospectively and include both sexes as well as community- and hospital-acquired bacteriuria. Novel risk factors (history of urogenital surgery, symptoms of hesitancy/retention, presence of the bacterial kpsMT gene) that could serve as clinical predictors and lead to the development of novel diagnostic tests and treatments were identified.
Notes
Acknowledgments. We thank Kyle Ota, who assisted with the collection of patient data, and Cherie Hill and Dorothy Sinclair for their invaluable help with data management. The study would also not have been possible without Joan Hoppe-Bauer and Linda Chase, who were instrumental in setting up the isolate identification and storage system. In Ann Arbor, Sreelatha Ponnaluri organized the isolate storage, and Mikiko Senga helped with DNA extraction in preparation for the microarray analyses.
Financial support. J. M. is supported by the National Institutes of Health (NIH) Clinical and Translational Science Award (CTSA; UL1RR024992) and is the recipient of a KL2 Career Development Grant (KL2RR024994). L. Z. is supported in part by a CTSA grant from the University of Michigan's Michigan Institute for Clinical & Health Research (UL1RR024986-01). B. F. is supported in part by an interdisciplinary training program grant at the University of Michigan (5T32AI049816-10) and by the National Institute of Child Health and Human Development (5R01HD038098-07) and the National Institute of Diabetes and Digestive and Kidney Diseases (5R21DK085290-02). D. K. W. was supported in part by a Prevention Epicenters Program grant from the Centers for Disease Control and Prevention(CDC 1U1CI000033301). J. P. H. is the recipient of a Burroughs-Wellcome Career Award for Medical Scientists and is supported by the NIH (HD001459-09 and DK064540-09).
Potential conflicts of interest. D. K. W. is a consultant for 3M Healthcare, Centene Corp, Bard, and Cardinal Health, and receives research funding from Sage Products, 3M Healthcare, bioMérieux, and Cubist Pharmaceuticals. J. M. has received payments for lectures from the Cape Girardeau Medical Society and the SHEA Education Committee.
All other authors report no potential conflicts.
All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.
References
- 1.Peralta G, Sanchez MB, Garrido JC, et al. Impact of antibiotic resistance and of adequate empirical antibiotic treatment in the prognosis of patients with Escherichia coli bacteraemia. J Antimicrob Chemother. 2007;60:855–63. doi: 10.1093/jac/dkm279. [DOI] [PubMed] [Google Scholar]
- 2.Olesen B, Kolmos HJ, Orskov F, Orskov I, Gottschau A. Bacteraemia due to Escherichia coli in a Danish university hospital, 1986–1990. Scand J Infect Dis. 1995;27:253–7. doi: 10.3109/00365549509019018. [DOI] [PubMed] [Google Scholar]
- 3.Melzer M, Petersen I. Mortality following bacteraemic infection caused by extended spectrum beta-lactamase (ESBL) producing E. coli compared to non-ESBL producing E. coli. J Infect. 2007;55:254–9. doi: 10.1016/j.jinf.2007.04.007. [DOI] [PubMed] [Google Scholar]
- 4.Johnson JR, Kuskowski MA, O'Bryan TT, Maslow JN. Epidemiological correlates of virulence genotype and phylogenetic background among Escherichia coli blood isolates from adults with diverse-source bacteremia. J Infect Dis. 2002;185:1439–47. doi: 10.1086/340506. [DOI] [PubMed] [Google Scholar]
- 5.Otto G, Sandberg T, Marklund BI, Ulleryd P, Svanborg C. Virulence factors and pap genotype in Escherichia coli isolates from women with acute pyelonephritis, with or without bacteremia. Clin Infect Dis. 1993;17:448–56. doi: 10.1093/clinids/17.3.448. [DOI] [PubMed] [Google Scholar]
- 6.Otto G, Magnusson M, Svensson M, Braconier J, Svanborg C. pap Genotype and P fimbrial expression in Escherichia coli causing bacteremic and nonbacteremic febrile urinary tract infection. Clin Infect Dis. 2001;32:1523–31. doi: 10.1086/320511. [DOI] [PubMed] [Google Scholar]
- 7.Ikaheimo R, Siitonen A, Karkkainen U, Mustonen J, Heiskanen T, Makela PH. Community-acquired pyelonephritis in adults: characteristics of E. coli isolates in bacteremic and non-bacteremic patients. Scand J Infect Dis. 1994;26:289–96. doi: 10.3109/00365549409011797. [DOI] [PubMed] [Google Scholar]
- 8.Moreno E, Planells I, Prats G, Planes AM, Moreno G, Andreu A. Comparative study of Escherichia coli virulence determinants in strains causing urinary tract bacteremia versus strains causing pyelonephritis and other sources of bacteremia. Diagn Microbiol Infect Dis. 2005;53:93–9. doi: 10.1016/j.diagmicrobio.2005.05.015. [DOI] [PubMed] [Google Scholar]
- 9.Venier AG, Talon D, Patry I, Mercier-Girard D, Bertrand X. Patient and bacterial determinants involved in symptomatic urinary tract infection caused by Escherichia coli with and without bacteraemia. Clin Microbiol Infect. 2007;13:205–8. doi: 10.1111/j.1469-0691.2006.01586.x. [DOI] [PubMed] [Google Scholar]
- 10.Tseng CC, Wu JJ, Liu HL, Sung JM, Huang JJ. Roles of host and bacterial virulence factors in the development of upper urinary tract infection caused by Escherichia coli. Am J Kidney Dis. 2002;39:744–52. doi: 10.1053/ajkd.2002.32992. [DOI] [PubMed] [Google Scholar]
- 11.Velasco M, Horcajada JP, Mensa J, et al. Decreased invasive capacity of quinolone-resistant Escherichia coli in patients with urinary tract infections. Clin Infect Dis. 2001;33:1682–6. doi: 10.1086/323810. [DOI] [PubMed] [Google Scholar]
- 12.Warren JW, Abrutyn E, Hebel JR, Johnson JR, Schaeffer AJ, Stamm WE. Guidelines for antimicrobial treatment of uncomplicated acute bacterial cystitis and acute pyelonephritis in women. Infectious Diseases Society of America (IDSA) Clin Infect Dis. 1999;29:745–58. doi: 10.1086/520427. [DOI] [PubMed] [Google Scholar]
- 13.Bone RC, Balk RA, Cerra FB, et al. Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. The ACCP/SCCM Consensus Conference Committee. American College of Chest Physicians/Society of Critical Care Medicine. Chest. 1992;101:1644–55. doi: 10.1378/chest.101.6.1644. [DOI] [PubMed] [Google Scholar]
- 14.Johnson JR, Jelacic S, Schoening LM, et al. The IrgA homologue adhesin Iha is an Escherichia coli virulence factor in murine urinary tract infection. Infect Immun. 2005;73:965–71. doi: 10.1128/IAI.73.2.965-971.2005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Yamamoto S. Molecular epidemiology of uropathogenic Escherichia coli. J Infect Chemother. 2007;13:68–73. doi: 10.1007/s10156-007-0506-y. [DOI] [PubMed] [Google Scholar]
- 16.Johnson JR, Kuskowski MA, Gajewski A, et al. Extended virulence genotypes and phylogenetic background of Escherichia coli isolates from patients with cystitis, pyelonephritis, or prostatitis. J Infect Dis. 2005;191:46–50. doi: 10.1086/426450. [DOI] [PubMed] [Google Scholar]
- 17.Guyer DM, Radulovic S, Jones FE, Mobley HL. Sat, the secreted autotransporter toxin of uropathogenic Escherichia coli, is a vacuolating cytotoxin for bladder and kidney epithelial cells. Infect Immun. 2002;70:4539–46. doi: 10.1128/IAI.70.8.4539-4546.2002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Johnson JR, Stell AL. Extended virulence genotypes of Escherichia coli strains from patients with urosepsis in relation to phylogeny and host compromise. J Infect Dis. 2000;181:261–72. doi: 10.1086/315217. [DOI] [PubMed] [Google Scholar]
- 19.Bauer RJ, Zhang L, Foxman B, et al. Molecular epidemiology of 3 putative virulence genes for Escherichia coli urinary tract infection—usp, iha, and iroN(E. coli) J Infect Dis. 2002;185:1521–4. doi: 10.1086/340206. [DOI] [PubMed] [Google Scholar]
- 20.Zhang L, Srinivasan U, Marrs CF, Ghosh D, Gilsdorf JR, Foxman B. Library on a slide for bacterial comparative genomics. BMC Microbiol. 2004;4:12. doi: 10.1186/1471-2180-4-12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Sandstedt SA, Zhang L, Patel M, et al. Comparison of laboratory-based and phylogenetic methods to distinguish between Haemophilus influenzae and H. haemolyticus. J Microbiol Methods. 2008;75:369–71. doi: 10.1016/j.mimet.2008.06.023. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Foxman B, Zhang L, Palin K, Tallman P, Marrs CF. Bacterial virulence characteristics of Escherichia coli isolates from first-time urinary tract infection. J Infect Dis. 1995;171:1514–21. doi: 10.1093/infdis/171.6.1514. [DOI] [PubMed] [Google Scholar]
- 23.Truzzi JC, Almeida FM, Nunes EC, Sadi MV. Residual urinary volume and urinary tract infection—when are they linked? J Urol. 2008;180:182–5. doi: 10.1016/j.juro.2008.03.044. [DOI] [PubMed] [Google Scholar]
- 24.May M, Brookman-Amissah S, Hoschke B, Gilfrich C, Braun KP, Kendel F. Post-void residual urine as a predictor of urinary tract infection—is there a cutoff value in asymptomatic men? J Urol. 2009;181:2540–4. doi: 10.1016/j.juro.2009.01.103. [DOI] [PubMed] [Google Scholar]
- 25.Lieber MM, Jacobsen SJ, Roberts RO, Rhodes T, Girman CJ. Prostate volume and prostate-specific antigen in the absence of prostate cancer: a review of the relationship and prediction of long-term outcomes. Prostate. 2001;49:208–12. doi: 10.1002/pros.1136. [DOI] [PubMed] [Google Scholar]
- 26.Johnson JR, Scheutz F, Ulleryd P, Kuskowski MA, O’Bryan TT, Sandberg T. Host-pathogen relationships among Escherichia coli isolates recovered from men with febrile urinary tract infection. Clin Infect Dis. 2005;40:813–22. doi: 10.1086/428048. [DOI] [PubMed] [Google Scholar]
- 27.Rijavec M, Muller-Premru M, Zakotnik B, Zgur-Bertok D. Virulence factors and biofilm production among Escherichia coli strains causing bacteraemia of urinary tract origin. J Med Microbiol. 2008;57:1329–34. doi: 10.1099/jmm.0.2008/002543-0. [DOI] [PubMed] [Google Scholar]
- 28.Bonacorsi S, Houdouin V, Mariani-Kurkdjian P, Mahjoub-Messai F, Bingen E. Comparative prevalence of virulence factors in Escherichia coli causing urinary tract infection in male infants with and without bacteremia. J Clin Microbiol. 2006;44:1156–8. doi: 10.1128/JCM.44.3.1156-1158.2006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Pavelka MS, Jr, Wright LF, Silver RP. Identification of two genes, kpsM and kpsT, in region 3 of the polysialic acid gene cluster of Escherichia coli K1. J Bacteriol. 1991;173:4603–10. doi: 10.1128/jb.173.15.4603-4610.1991. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Robbins JB, McCracken GH, Jr, Gotschlich EC, Orskov F, Orskov I, Hanson LA. Escherichia coli K1 capsular polysaccharide associated with neonatal meningitis. N Engl J Med. 1974;290:1216–20. doi: 10.1056/NEJM197405302902202. [DOI] [PubMed] [Google Scholar]
- 31.Kim BY, Kang J, Kim KS. Invasion processes of pathogenic Escherichia coli. Int J Med Microbiol. 2005;295:463–70. doi: 10.1016/j.ijmm.2005.07.004. [DOI] [PubMed] [Google Scholar]
- 32.Cegelski L, Marshall GR, Eldridge GR, Hultgren SJ. The biology and future prospects of antivirulence therapies. Nat Rev Microbiol. 2008;6:17–27. doi: 10.1038/nrmicro1818. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Goller CC, Seed PC. High-throughput identification of chemical inhibitors of E. coli group 2 capsule biogenesis as anti-virulence agents. PLoS One. 2010;5:e11642. doi: 10.1371/journal.pone.0011642. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Russo TA, Beanan JM, Olson R, MacDonald U, Cope JJ. Capsular polysaccharide and the O-specific antigen impede antibody binding: a potential obstacle for the successful development of an extraintestinal pathogenic Escherichia coli vaccine. Vaccine. 2009;27:388–95. doi: 10.1016/j.vaccine.2008.10.082. [DOI] [PubMed] [Google Scholar]
- 35.Vigil PD, Stapleton AE, Johnson JR, et al. Presence of putative repeat-in-toxin gene tosA in Escherichia coli predicts successful colonization of the urinary tract. MBio. 2011;2:e00066. doi: 10.1128/mBio.00066-11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Henderson JP, Crowley JR, Pinkner JS, et al. Quantitative metabolomics reveals an epigenetic blueprint for iron acquisition in uropathogenic Escherichia coli. PLoS Pathog. 2009;5:e1000305. doi: 10.1371/journal.ppat.1000305. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Hsu CY, Fang HC, Chou KJ, Chen CL, Lee PT, Chung HM. The clinical impact of bacteremia in complicated acute pyelonephritis. Am J Med Sci. 2006;332:175–80. doi: 10.1097/00000441-200610000-00004. [DOI] [PubMed] [Google Scholar]
- 38.Chen Y, Nitzan O, Saliba W, Chazan B, Colodner R, Raz R. Are blood cultures necessary in the management of women with complicated pyelonephritis? J Infect. 2006;53:235–40. doi: 10.1016/j.jinf.2005.12.005. [DOI] [PubMed] [Google Scholar]
- 39.Velasco M, Martinez JA, Moreno-Martinez A, et al. Blood cultures for women with uncomplicated acute pyelonephritis: are they necessary? Clin Infect Dis. 2003;37:1127–30. doi: 10.1086/378291. [DOI] [PubMed] [Google Scholar]
- 40.Jerkeman M, Braconier JH. Bacteremic and non-bacteremic febrile urinary tract infection—a review of 168 hospital-treated patients. Infection. 1992;20:143–5. doi: 10.1007/BF01704603. [DOI] [PubMed] [Google Scholar]