ABSTRACT
In the face of increasing rates of antimicrobial resistance in complicated urinary tract infections (cUTIs), clinicians need to understand cross-resistance patterns among commonly encountered pathogens. We performed a multicenter, retrospective cohort study in the Premier database of approximately 180 hospitals, from 2013 to 2018. Using an ICD-9/10-based algorithm, we identified all adult patients hospitalized with cUTIs and included those with a positive blood or urine culture. We examined the microbiology and susceptibilities to common cUTI antimicrobials (3rd-generation cephalosporin [C3], fluoroquinolones [FQ], trimethoprim-sulfamethoxazole [TMP/SMZ], fosfomycin [FFM], and nitrofurantoin [NFT]) singly and in groups of two. Among 28,057 organisms from 23,331 patients, the 3 most common pathogens were Escherichia coli (41.0%; C3r, 15.1%), Klebsiella pneumoniae (12.1%; C3r, 13.2%), and Pseudomonas aeruginosa (11.0%; C3r, 12.0%). E. coli was most frequently resistant to FQ (43.5%) and least to NFT (6.7%). K. pneumoniae was most frequently resistant to NFT (60.8%) and least to FFM (0.1%). P. aeruginosa was most frequently resistant to FQ (34.4%) and least to TMP/SMZ (4.2%). Of the C3r E. coli isolates, 87.1% were also FQr, 63.7% were TMP/SMZr, and 13.3% were NFTr. C3r K. pneumoniae isolates had a 76.5% chance of being FQr, 78.1% were TMP/SMZr, and 77.6% were NFTr. C3r P. aeruginosa coexisted with FQr in 47.3%, TMP/SMZr in 18.9%, and NFTr in 28.7%. Among the most common pathogens isolated from hospitalized patients with cUTIs, the rates of single resistance to common treatments and of cross-resistance to these regimens are substantial. Knowing the patterns of cross-resistance may help clinicians tailor empirical therapy more precisely.
INTRODUCTION
Urinary tract infections (UTIs) in adults account for more than 400 million annual hospitalizations in the United States (1). Moreover, the prevalence of such admissions has nearly doubled between 1998 and 2011 (1). Although during this time, the associated hospital length of stay (LOS) has decreased by 1 day (on average), the median attributable costs of these hospitalizations have doubled from $2,400 to $5,000 over the same time period (1). In short, UTIs represent a major strain on the health care system.
By its very definition, uncomplicated UTI—one that occurs in essentially healthy females with no evidence of systemic involvement—comprises a minority of all UTIs that require hospitalization. The vast majority of hospitalizations for UTIs are considered complicated (cUTI). As such, patients who require inpatient treatment for their cUTIs likely present more challenging treatment dilemmas. Specifically, these patients, in addition to their underlying comorbidities, face a higher risk of harboring a pathogen resistant to commonly employed therapies (2–4).
Of particular concern in the care of patients with cUTIs are increasing rates of resistance to commonly used first-line agents, including 3rd-generation cephalosporins (C3), fluoroquinolones (FQ), and trimethoprim-sulfamethoxazole (TMP/SMZ) (2–5). Because of the frequency of cUTI diagnosis, coupled with escalating rates of antimicrobial resistance (AMR), carbapenems are being more frequently prescribed as first-line therapy for cUTI. This behavior by physicians, although undertaken to ensure the patient receives initially appropriate antibiotic therapy, creates increased selection pressure and fuels further AMR.
One potential way to limit the use of overly broad agents is to understand better the microbiology as it impacts specific infectious syndromes. Pneumonia and cUTI, for example, differ substantially with respect to their pathogen distributions. That is, while Escherichia coli is the most common cause of cUTI, enteric pathogens are rarely causative in pneumonia (6). For this reason, when considering empirical treatment in the setting of cUTI, AMR among E. coli isolates becomes much more salient than in the case of pneumonia. With the rise in AMR, however, there is also an increase in multiply resistant pathogens, which may make empirical choices even trickier. In other words, it would be useful for clinicians to know whether resistance to a single agent of interest may be a marker for resistance to another drug aimed at the same pathogen. To explore patterns of resistance in cUTI to multiple agents, we examined the microbiology and in vitro susceptibilities to commonly used antimicrobial regimens among patients hospitalized with cUTIs, singly and in combination.
(Data from this study have in part been presented at ID Week 2019.)
RESULTS
Study enrollment is depicted in Fig. 1. Among 23,331 patients meeting enrollment criteria, 28,057 organisms of interest were isolated. The vast majority (68.3%) of the isolates cooccurred in blood and urine, while 31.0% were found in urine only. Isolates from blood only comprised <1% of all pathogens. (For hospital characteristics, see Table S3 in the supplemental material.)
Study enrollment. cUTI, complicated urinary tract infections; LOS, length of stay; cIAI, complicated intraabdominal infection.
The 3 leading pathogens were Escherichia coli, Klebsiella pneumoniae, and Pseudomonas aeruginosa, which together comprised nearly two-thirds of all isolates (Table 1). The 10 most commonly isolated organisms accounted for more than 90% of all cUTIs.
Organism distribution
Organisms varied with respect to the frequency of susceptibility testing to different antimicrobials: FQ susceptibility was the most and fosfomycin (FFM) the least frequently tested for (Fig. 2; see also Table S4). Of the 3 most commonly isolated cUTI organisms, resistance to C3 was found in 15.1% of E. coli isolates, 13.2% for K. pneumoniae, and 12.0% for P. aeruginosa (Table 2). E. coli was most frequently resistant to FQ (43.5%) and least frequently resistant to nitrofurantoin (NFT) (6.7%). It was never tested for susceptibility to FFM. K. pneumoniae was most frequently resistant to NFT (60.8%) and least frequently resistant to FFM (0.1%), the latter likely due to FFM testing being a rare event (Table S4). P. aeruginosa was most frequently resistant to FQ (34.4%) and least frequently resistant to TMP/SMZ (4.2%). Similarly to E. coli, it was never tested for susceptibility to FFM. NFT and FQ exhibited some of the highest rates of resistance across multiple cUTI pathogens (Table 2).
Frequencies of susceptibility testing. C3, 3rd-generation cephalosporin; FQ, fluoroquinolone; TMP/SMX, trimethoprim-sulfamethoxazole; FFM, fosfomycin; NFT, nitrofurantoin.
Prevalence of resistance among 10 most common cUTI pathogens
Table 3 shows the patterns of cross-resistance among the 10 most common cUTI pathogens to combinations of two of the examined antimicrobial regimens. There was a striking overlap between C3r and FQr. Focusing on the most prevalent cUTI pathogen, of the C3r E. coli isolates, 87.1% were also FQr, 63.7% were TMP/SMZr, and 13.3% were NFTr. Conversely, just less than one-third of all FQr E. coli isolates were C3r as well. Similar proportions of C3r E. coli isolates were also TMP/SMZr and NFTr (Table 3). NFT was the drug E. coli isolates were least likely to exhibit cross-resistance to in the presence of resistance to another antimicrobial. (For resistance trends over time, see Table S5a to e).
Cross-resistance among 10 most common cUTI pathogensa
The prevalence of cross-resistance to C3 and other antimicrobials was high in K. pneumoniae. For example, when C3r, K. pneumoniae had a 76.5% chance of being FQr, 78.1% of being TMP/SMZr, and 77.6% of being NFTr. With the exception of NFTr K. pneumoniae isolates, where cross-resistance to C3, FQ, and TMP/SMZ was <30%, and FFMr, which was rarely tested, the chances of cross-resistance to all the remaining drugs were >50% across the board and, in the case of NFTr, among FQr isolates, as high as 81.9% (Table 3).
As for P. aeruginosa, when C3r was present, FQr coexisted in 47.3% of the isolates compared to relatively lower likelihoods of TMP/SMZr (18.9%) and NFTr (28.7%). Among the pool of FQr P. aeruginosa isolates, the rate of resistance to other therapies was <20% across the board and lowest to TMP/SMZ (5.1%). TMP/SMZr, however, was associated with a high risk (>40%) of resistance to other drugs, as was NFTr (>35%) (Table 3). The patterns of cross-resistance among other less common pathogens are depicted in Table 3.
DISCUSSION
We have demonstrated that the rates of single and cross-resistance to common cUTI regimens used to treat the leading cUTI pathogens are significant: the rate of single TMP/SMZr was >36%, and the rate of FQr exceeded 44%. Although C3r is relatively less common than FQr, its presence signals a high risk for FQr and TMP/SMZr among E. coli isolates, the pathogen that accounts for >40% of all cUTIs. This tight association between C3r and resistance to other frequently utilized anti-UTI therapies has implications for C3r E. coli at the hospital level. That is, institutions with a high prevalence of C3r are likely to have commensurately high levels of resistance to FQ and TMP/SMZ. Even NFTr, though lower than others in the presence of C3r among E. coli isolates, carries a risk of 10%. Similar conclusions can be drawn for K. pneumoniae. When a hospital harbors a high volume of K. pneumoniae isolates resistant to one common antimicrobial, the levels of resistance to others can also be expected to be high, or >50% for C3, FQ, and TMP/SMZ, and lower, albeit still substantial, for NFT. In the case of P. aeruginosa, institutions with a high prevalence of TMP/SMZr or NFTr are also at risk for resistance to the remaining regularly used agents. Put simply, with the current rates of resistance generally encountered in cUTIs, one cannot presume that if a pathogen is resistant to a specific agent that another option will be readily available. This pattern is consistent across a range of pathogens and range of treatment alternatives.
Emerging antimicrobial resistance has complicated clinicians’ choices for empirical therapy in serious infections requiring hospitalization. In many of these infections, when appropriate therapy that covers the presumed pathogen is delayed, outcomes for patients are worse in both clinical and economic terms (7–12). Yet AMR, when present, contributes significantly to the risk of receiving inappropriate empirical coverage (12–14). Because empirical coverage is, by definition, required before definitive culture results become available, a probabilistic approach, in which clinicians make a judgment as to what represents their most reliable option, is necessary. Awareness of the local microbiology and susceptibilities, therefore, can lead to a more informed choice of therapy. Thus, our observations on what resistance patterns tend to occur together is unique in that it is tailored to help clinicians gain a deeper understanding of how to choose empirical treatments. Furthermore, our data analysis facilitates decision making conditioned not only on patterns of resistance to a single drug but also on the risk of additive resistance to commonly prescribed antimicrobial regimens.
It is alarming that the top three non-carbapenem-resistant pathogens responsible for culture-positive cUTIs exhibit high rates of AMR to such common treatments as C3, FQ, and TMP/SMZ. Among those three agents, with the exception of TMP/SMZr P. aeruginosa, all other rates of resistance exceeded the 10% threshold beyond which formal treatment guidelines discourage empirical use (2). Our findings generally confirm the observations of others in that our estimated rates of resistance are similar to those reported in the literature for select pathogens. Additionally, we add to the observations of others by analyzing a more comprehensive group of bacterial isolates (15, 16). It is further important to appreciate that AMR frequently does not occur singly, implying that many of the drugs clinicians rely upon in treating cUTI may be suboptimal choices. However, knowing which groups of AMR travel together, along with an appreciation of local antibiogram patterns, may aid physicians in making more informed treatment decisions.
It is worth noting that fosfomycin susceptibility testing was exceedingly rare. There are likely several reasons for it. First, it is an old drug whose use waned with the introduction of newer agents. Second, given its infrequent use in recent decades, there are no standard breakpoints for fosfomycin in the majority of UTI pathogens (17). Third, since automated fosfomycin testing is not available, it is possible that the frequency of manual testing is underreported. However, its broad spectrum and pharmacodynamics make it potentially attractive as a UTI treatment in the face of surging AMR. Indeed, in some studies, the susceptibility rates of common UTI pathogens to fosfomycin are 90% (18).
Our study has a number of strengths and limitations. Premier, as a large multihospital database, does not suffer from lack of generalizability. As an observational study, however, it is prone to selection bias; we attempted to mitigate its magnitude by defining the cohort prospectively. Misclassification is an issue in any study and particularly when using administrative data. To minimize it, we used a previously published algorithm to identify our cohort. However, this algorithm has not been clinically validated (19). Though there are multiple ways to define cUTI, our definition aligns closely with clinical practice. We developed a restrictive definition for cUTI in order to optimize its specificity. This may have reduced its sensitivity, resulting in the exclusion of at least some cUTI cases. Furthermore, we excluded other potential sources of infection and included microbiology specimens, pharmacy data, and dates of cultures and treatments. This is a descriptive analysis, and adjustment for confounding was not undertaken. In some cases, we report data for antibiotic pathogen susceptibility testing which would either not normally be performed or, if performed, not be reported. This may seem to limit the validity of our findings. While these data do not reflect recommendations for specific testing among these antibiotic-pathogen combinations, we included the results in our paper to illustrate how automated susceptibility testing is currently done in many U.S. hospitals.
In summary, we have demonstrated that, among frequently isolated bacterial pathogens in the setting of hospitalizations with cUTI, resistance to common antimicrobial treatments is high. Moreover, where resistance to a single drug or class is detected, there are variable, but uniformly substantial, risks of resistance to additional drugs in the arsenal used in cUTIs. In addition to considering local antibiograms, understanding the combinations of resistance patterns among common pathogens to common antimicrobial regimens may help target better empirical treatment choices.
MATERIALS AND METHODS
Ethics statement.Because this study used already existing fully deidentified data, it was exempt from IRB review under US 45 CFR 46.101(b)4 (20).
Study design and patient population.We conducted a multicenter, retrospective cohort study of hospitalized patients with culture-positive carbapenem-susceptible cUTI to explore the prevalence and impact of resistance to commonly used noncarbapenem empirical regimens. The case identification approach (see Table S1 in the supplemental material) relied on a previously published algorithm (19). Briefly, we included all adult patients (age, ≥18 years) with a urine culture obtained at any time during hospitalization who received antibiotic treatment on the day of the index culture and continued for ≥3 consecutive days and who met the definition for cUTI (19). Only culture-positive (urine, blood, or both) patients were included in the cohort. Patients with a hospital length of stay (LOS) of <2 days, who fit the definition for a complicated intraabdominal infection (to reduce the risk of misclassification) (see Table S2), who were transferred from another acute care facility, or who grew an organism resistant to at least one carbapenem were excluded (19).
Data source.The data for the study were derived from Premier Research database, an electronic laboratory, pharmacy, and billing data repository for years 2013 through 2018. The database has been described in detail previously (19, 21–23). We used data from a subset of approximately 180 U.S. institutions (of >900 hospitals submitting patient data to Premier) who submitted microbiology data annually during the study time frame.
Microbiology and susceptibilities.Organisms were classified as susceptible (S), intermediate (I), or resistant (R). For the purposes of the current analyses, I and R were grouped together as resistant. For each of the common antimicrobials of interest (C3, FQ, TMP/SMZ, fosfomycin [FFM], and nitrofurantoin [NFT]), we examined specific susceptibility testing to determine the resistance status. Cross-resistance was defined as the risk of resistance to a second antimicrobial if resistance to a single antimicrobial was detected. The results of susceptibility testing represented the findings of the local microbiology laboratories at the participating hospitals.
The first detection of an organism served as the index culture. To be considered culture positive, a qualifying common bacterium had to grow from urine or blood samples from the patient. Organisms of interest included Enterobacteriaceae, P. aeruginosa, Acinetobacter baumannii, Enterococcus faecium, Enterococcus faecalis (19). All microbiology results were based on the local testing done by participating hospitals.
Outcome variables and statistical analyses.We examined the prevalence of antimicrobial resistance singly and cross-resistance between groups of two of the drugs/classes of interest. No hypothesis testing was undertaken. Only descriptive statistics are presented.
Data availability.The data used in this study derive from Premier Research database, a proprietary third-party database available to researchers through a specific agreement with Premier.
ACKNOWLEDGMENTS
This study was supported by a grant from Spero Therapeutics, Cambridge, MA, USA.
M.D.Z. is a consultant to Spero Therapeutics. Her employer, EviMed Research Group, LLC, has received research grant support from Spero Therapeutics. B.H.N.’s employer, OptiStatim, LLC, has received support from EviMed Research Group, LLC. K.S. is an employee of and stockholder in Spero Therapeutics. A.F.S. is a consultant to and has received research grant support from Spero Therapeutics. M.D.Z. and A.F.S. have received grant support and/or have served as consultants to Merck, Melinta, Tetraphase, Pfizer, Astellas, Shionogi, The Medicines Company, Lungpacer, and Theravance.
M.D.Z., K.S., and A.F.S. contributed substantially to the study design, data interpretation, and the writing of the manuscript. B.H.N. had full access to all of the data in the study, takes responsibility for the integrity of the data and the accuracy of the data analysis, and contributed substantially to the study design, data analysis, and the writing of the manuscript. No persons other than the authors participated in the study or the writing of the manuscript.
FOOTNOTES
- Received 21 February 2020.
- Returned for modification 4 April 2020.
- Accepted 26 April 2020.
- Accepted manuscript posted online 18 May 2020.
Supplemental material is available online only.
- Copyright © 2020 American Society for Microbiology.
