Patient characteristics but not virulence factors discriminate between asymptomatic and symptomatic E. coli bacteriuria in the hospital

Escherichia coli is the most frequent pathogen to cause symptomatic urinary tract infection (sUTI) but can also lead to asymptomatic bacteriuria. Asymptomatic bacteriuria (ASB) denotes bacterial colonization of the urogenital tract without subjective or systemic host responses. ASB screening and treatment are recommended only during pregnancy [1] and in the preoperative evaluation of men before urological procedures [2], circumstances where preemptive antibiotic administration decreases the risk of infectious complications [3]. Despite these recommendations, overuse of antibiotics for ASB is common and clearance of bacteriuria is often transient, leading to further treatment courses [4]. Inappropriate antibiotics result in increased healthcare costs and foster antimicrobial resistance [5]; they can even eliminate the protective effect that ASB may have against recurrent UTIs [6]. These issues have led to calls for assessing hospitals’ performance in reducing inappropriate antibiotic use nationwide [7].

Surprisingly few studies have compared patients with symptomatic bacteriuria to patients with ASB to identify patient-level risk factors for symptomatic presentation or differences in bacterial virulence factors [8‐12]. Results from these studies have been inconsistent, reporting either a variety of virulence factors associated with symptomatic bacteriuria [8‐11] or no association [12]. We found no previous studies addressing clinical risk factors. Urine cytokines such as interleukin 6 and 8 may be lower in ASB in certain populations [13, 14], but an objective test to discriminate ASB from symptomatic bacteriuria is not commercially available. A better understanding of what differentiates ASB from symptomatic bacteriuria could be particularly helpful in patients who are unable to report symptoms, such as intubated patients or those with altered mental status. Here, we used a prospective cohort of patients with E. coli bacteriuria as identified by clinicians in a hospital setting. The purpose of our study was to characterize host and pathogen factors associated with E. coli ASB that could subsequently be used in predictive models, lead to the development of bedside tests to distinguish ASB from sUTI, or guide treatment decision-making.

We identified 337 patients with E. coli bacteriuria during the study period. Of these, 50 (14.8%) were excluded because they were considered unclassifiable (e.g., intubated patients or patients with acutely altered mental status who were unable to report symptoms). Among the remaining 287 patients, 110 (38.3%) were classified as having asymptomatic bacteriuria, and 177 (61.7%) as having symptomatic UTI. Two hundred and twenty-five patients were female (78.4%), and 169 (58.9%) were white (Table 1). Age had a bimodal distribution, with a small peak between 20 and 25 years and a larger peak between 60 to 80 years; the median age was 65 years (range, 19–101). Among symptomatic patients, 70 (39.4%) had cystitis, and 107 (61.6%) pyelonephritis. In this group of symptomatic patients, 91 (51.4%) were tested for bloodstream infection, and 20 had positive blood cultures. Among asymptomatic patients, 35 (31.8%) had blood cultures taken, and one was positive for E. coli.

Why E. coli bacteriuria is associated with urinary tract symptoms in some instances but not in others remains incompletely understood. This knowledge gap may explain why E. coli bacteriuria is often treated with antibiotics independent of any symptoms [19]. Overtreatment in turn generates added cost and can facilitate the development of antibiotic resistance. It would therefore be invaluable if we had tools to differentiate between symptomatic infection and asymptomatic bacteriuria (and to understand which bacteriurias result in morbidity in patients who are unable to report symptoms). Here, we report a comparison of patient characteristics and virulence factors between symptomatic and asymptomatic patients with E. coli bacteriuria. While we found independent clinical predictors for asymptomatic bacteriuria (chronic pulmonary disease, dementia) no virulence pattern could be associated with the corresponding E. coli isolates.

To our knowledge there are no published studies of patient characteristics to differentiate ASB from symptomatic UTI. We found the comorbidities chronic pulmonary disease and dementia to be associated with asymptomatic bacteriuria. It is conceivable that underlying dementia resulted either in poor symptom recognition by the affected patients or may have impaired their ability to report symptoms. ASB is also much more prevalent in elderly long-term care residents [20] who are prone to cognitive deficits, and cognitive changes may be a risk factor for ASB [21]. Other potential contributors are poor perineal hygiene and incomplete bladder emptying. It is less clear why chronic pulmonary disease should be associated with ASB; this may be due to different screening practices in medical subspecialties or a spurious finding. Those with ASB may finally represent a group of patients in which it was more difficult to find a diagnosis and urine cultures were part of the broad work-up.

A small number of studies have determined differences in bacterial virulence factors between strains causing ASB and sUTI. Mabbett et al. compared 57 ASB strains to 45 cystitis/pyelonephritis strains not only in regard to their virulence genes but also in gene product expression and epithelial adhesion studies [8]. PapGII (47% vs. 16%) and afa (29% vs. 11%) were associated with sUTI as were hemolysin expression, siderophore expression, and adhesion to epithelial cells. However, the authors noted that ASB and sUTI strains tended to be phylogenetically related. The latter finding was confirmed by Takahashi et al. who investigated virulence factors in 283 E. coli urinary isolates from a hospital in Japan, and also saw a similar distribution of serotypes across ASB and sUTI isolates. The virulence genes iha (31% vs. 13%), ompT (82% vs. 70%) and PAI, which refers to the so-called genomic “pathogenicity island” (65% vs. 50%), were all significantly more frequent in strains causing sUTI [9]. Although ompT was a common virulence gene in our similar-sized cohort, the sUTI strains we analyzed did not carry ompT more frequently (83% vs. 76%). A number of smaller studies in subpopulations like pregnant women and patients with spinal cord injuries found other virulence factors to be associated with symptomatic infections [10, 11] or were negative [22]. In addition to the bacterial characteristics mentioned above, there have been reports that ASB strains can be identified based upon their capability to form biofilms in vitro[23], for which we did not screen our isolates. With 287 patient samples our cohort is one of the largest to date; however, we did not find any virulence gene to predominate in the subset of symptomatic patients. One of the possible explanations for this negative finding is that virulence genes in ASB may carry deletions that block gene expression [24‐26]; therefore, ASB strains may look more virulent by genotype than they actually are in vivo. Another possibility is that virulence factors other than those we examined are more relevant in determining pathogenicity [27]. Yet another possibility is that the major determinant of symptom development is unrelated to pathogen factors, but due to unmeasured factors such as the host’s innate immune response [28‐30]. Future research should incorporate studies of the host’s defense mechanisms (e.g., toll-like receptor polymorphisms) into a multi-faceted analysis of symptom development in bacteriuria.

There are a number of limitations to our study. Including only patients that were tested for and found to have bacteriuria may have introduced selection bias. We suspect that this bias was particularly relevant for ASB patients which were included despite an indication for culturing. Symptoms at the time of bacteriuria were taken from hospital charts entered by physicians and nurses; therefore patient symptoms may have been underreported. We did not obtain urine samples to determine whether virulence factors were expressed in vivo, nor test gene expression of the bacterial isolates in vitro. Also, we elected to test for a specific set of virulence factors but may have missed others that are not identified yet. Blood cultures were ordered by the treating physicians and may have missed septic or febrile patients at risk for bacteremia. Lastly, we did this study at a single tertiary-care center and the findings may not be generalizable to other settings. This also goes for the high levels of resistance to the most commonly used antimicrobials that we encountered here.