Comparative Analysis of Complicated Urinary Tract Infections Caused by Extensively Drug-Resistant Pseudomonas aeruginosa and Extended-Spectrum β-Lactamase-Producing Klebsiella pneumoniae

The objective was to compare clinical characteristics, outcomes, and economic differences in complicated urinary tract infections (cUTI) caused by extensively drug-resistant Pseudomonas aeruginosa (XDR P. aeruginosa) and extended-spectrum beta-lactamase-producing Klebsiella pneumoniae (ESBL-K. pneumoniae). A retrospective study was conducted at a tertiary care hospital. Patients with XDR P. aeruginosa and ESBL-K. pneumoniae cUTIs were compared. The primary outcome was clinical failure at day 7 and at the end of treatment (EOT). Secondary outcomes: 30- and 90-day mortality, microbiological eradication, and economic cost. Two-hundred and one episodes were included, of which 24.8% were bloodstream infections. Patients with XDR P. aeruginosa cUTI more frequently received inappropriate empirical therapy (p < 0.001). Nephrotoxicity due to antibiotics was only observed in the XDR P. aeruginosa group (26.7%). ESBL-K. pneumoniae cUTI was associated with worse eradication rates, higher recurrence, and higher infection-related readmission. In multivariate analysis, XDR P. aeruginosa was independently associated with clinical failure on day 7 of treatment (OR 4.34, 95% CI 1.71–11.04) but not at EOT, or with mortality. Regarding hospital resource consumption, no significant differences were observed between groups. XDR P. aeruginosa cUTI was associated with worse early clinical cures and more antibiotic side effects than ESBL-K. pneumoniae infections. However, no differences in mortality or in hospitalization costs were observed.


Introduction
Pseudomonas aeruginosa has an extraordinary ability to develop resistance through chromosomal mutations and the acquisition of resistance genes [1][2][3]. The emergence of infections caused by multidrug-resistant and extensively drug-resistant Pseudomonas aeruginosa (MDR P. aeruginosa and XDR P. aeruginosa) has become a serious global concern [4]. the first 48 h of the episode or were lost to follow-up were also excluded. This paper was written following the STROBE guidelines for observational studies.
UTI episodes caused by XDR P. aeruginosa from January 2010 to June 2019 were included in the XDR P. aeruginosa cohort. This cohort was previously analyzed to evaluate the efficacy and safety of aminoglycosides or polymyxin monotherapy compared with other antibiotic regimens [27]. XDR P. aeruginosa UTI episodes were matched 1:1 with cases of ESBL-K. pneumoniae. UTI ESBL-K. pneumoniae episodes were consecutively included from February 2013 until the required sample size was reached November 2019. In total, 201 patients were included: 101 patients in the XDR P. aeruginosa cohort and 100 in the ESBL-K. pneumoniae cohort. Patients were followed for up to 90 days from the date of the urine culture. In cases of more than one episode of XDR P. aeruginosa or ESBL-K. pneumoniae UTI, the second and following episodes were assessed if they occurred at least 90 days after the prior one.

Clinical Variables, Data Sources, and Definitions
Three authors (E.S., I.L.M. and S.G.-Z.) were responsible for collecting all data from the hospital's electronic medical records. Clinical and demographic data were recorded, including age, sex, underlying conditions such as diabetes mellitus, chronic obstructive pulmonary disease (COPD), congestive heart failure, cirrhosis, neurological disorder, hematologic and solid tumor malignancy, as well as the Charlson comorbidity severity index [28]. Finally, the existence of a nephro-urological history was also recorded, including chronic kidney disease, dialysis, renal transplant, urological neoplasia, prior history of benign prostatic hypertrophy, obstructive urinary disease, recurrent UTI, and indwelling urinary catheter or other urological devices in the preceeding 14 days.
With respect to UTI classification, acute pyelonephritis was considered if the patient had at least two of the following criteria: a temperature above 37.7 • C, UTI symptoms (dysuria, urgency, suprapubic pain, and/or pollakiuria), local pain (lumbar back pain and/or pelvic or perineal pain), and/or altered mental status in persons up to 70 years of age. Complicated UTI was established in those patients with the same criteria and a prior history of benign prostatic hyperplasia, intermittent or permanent indwelling urinary catheter (or removal within 48-72 h before the onset of infection), or underlying urological abnormalities such as nephrolithiasis, stricture, stents, history of renal transplantation, urinary diversion or neurogenic bladder.
Acquisition of infection was classified in accordance with Friedman et al. [29] as community-acquired, healthcare-related, or nosocomial. Healthcare-associated risk factors, such as antibiotic exposure, hospital stay, surgery, or ICU admission in the previous three months, and/or residence in a long-term care facility, were also collected.
Baseline disease severity was measured by the Sequential Organ Failure Assessment (SOFA) score [30], the Quick SOFA score (qSOFA) [31], the Simplified Acute Physiology Score (SAPS II) [32], the need for intensive care unit (ICU) admission and the presence of sepsis or septic shock [31]. In the case of bacteraemia, the Pitt score was also calculated [32].
Recurrence was defined as repeated signs or symptoms of UTI and a urinary isolate with the same susceptibility profile as the index infection. Reinfection was defined as recurrent signs or symptoms of UTI caused by a different strain than the index infection strain. Microbiological eradication was considered if there was no growth of XDR P. aeruginosa or ESBL-producing K. pneumoniae in the final urine culture, if available. Episodes with missing urine samples during follow-up were classified as indeterminate.
In terms of management, appropriate empirical and definitive antibiotic treatments were assessed. Appropriate antibiotic therapy was considered when at least one antibiotic administered displayed documented in vitro susceptibility in accordance with the breakpoints established by the European Committee on Antimicrobial Susceptibility Testing (EUCAST) [33,34]. Adequate source control was also collected, defined as removal or insertion of an indwelling urinary catheter, percutaneous drainage of the urinary tract (double J stent, nephrostomy), or surgical intervention, as appropriate.

Outcomes and Follow-Up
The primary outcome was clinical failure, determined at day 7 and at the end of treatment (EOT). Clinical failure was considered if there was persistence or worsening of signs and/or symptoms of UTI, modification of the antibiotic therapy due to side effects, and/or death.
Secondary outcomes were 30-day and 90-day mortality, recurrence, microbiological eradication, percentage of infection-related readmissions, and economic cost per episode. Finally, antibiotic-induced side effects were also evaluated. The follow-up period was 90 days from the date of the first urine culture.

Microbiological Studies
Urine cultures were performed as part of the clinical routine, following standard laboratory procedures. Cultures with growth yielding ≥10 5 colony-forming units/mL of a single bacterial type in a urine sample (collected midstream) were considered positive. Bacterial identification was performed by conventional biochemical tests and matrix-assisted laser-desorption and ionization time-of-flight mass spectrometry (MALDI-TOF MS), using a Bruker Microflex ® LT instrument and the MALDI Biotyper ® software (Bruker Daltonics, MA, USA). Antibiotic susceptibility testing (AST) of isolates was performed by broth microdilution, using MicroScan panels [Beckman-Coulter] on the automated MicroScan WalkAway system [Beckman-Coulter]. Results were interpreted according to the European Committee for Antimicrobial Susceptibility Testing (EUCAST) guidelines. The following antimicrobials were tested in both species: ceftazidime, cefepime, piperacillintazobactam, imipenem, meropenem, aztreonam, ciprofloxacin, gentamicin, tobramycin, amikacin. Amoxicillin-clavulanic acid, cefuroxime, trimethoprim/sulfamethoxazole, and nitrofurantoin, were also tested in K. pneumoniae; colistin was assessed in P. aeruginosa. Ceftolozane-tazobactam was not routinely used for a large part of the study; it was tested by the gradient diffusion method (Etest, bioMérieux, Marcy-l'Etoile, France) [35] from 2017 onwards. The ESBL screening was based on the AST results of third-generation cephalosporins (MIC > 1 mg/L for cefotaxime/ceftriaxone and ceftazidime) following EUCAST guidelines [29,30]. ESBL production was further confirmed phenotypically by the MIC reduction of any of these cephalosporins combined with clavulanic acid compared with the MIC of the same cephalosporin alone in the microdilution, or by the double-disk diffusion test (DDST) [29,30]. XDR phenotype in P. aeruginosa was defined according to Magiorakos et al. [36].

Statistical Analysis
The sample size required (101 per group) was determined from the results of previous studies [19,23] to detect a difference of almost 20% in the 7-day clinical response between XDR P. aeruginosa or ESBL-K. pneumoniae UTI; statistical power was set at 80%, alpha error at 0.05, and the estimated loss to follow-up at 25%.
Categorical variables were presented as numbers and percentages and were compared by the X 2 test or Fisher's exact test. Continuous variables were expressed as the median and interquartile range (IQR) and compared by Student's t-test or the Mann-Whitney U test, as appropriate. Stepwise logistic regression with variable selection was used to examine independent variables associated with clinical failure (both at day 7 of treatment and at end of treatment). Variables with p value < 0.20 in the unadjusted analysis, as well as those that were clinically relevant but not statistically significant, were included in the model. Results were expressed as odds ratios (OR) and its 95% confidence interval (95% CI). An adjusted Cox regression model was used to assess 30-day and 90-day mortality. Variables with p value < 0.20 in the unadjusted analysis and/or those considered clinically relevant were included in the model. Results were expressed as hazard ratio (HR) and 95% CI. Finally, unadjusted and adjusted analysis of hospitalization costs were performed through median regression models to assess variables associated with significantly higher costs. The choice of median regression, rather than linear regression, was due to the fact that the distribution of costs was highly skewed to the right (and so did not meet the assumption of normality); p < 0.05 was considered statistically significant and all analyses were two-tailed. Statistical analyses were carried out with SPSS Statistics 26.0 software.

Cost Estimation
The Municipal Institute of Health uses a hospital cost accounting system based on full-cost allocation to estimate the direct costs of clinical activity. In the present study, cost estimation was based on a full costing method and the criteria of clinical activity-based costing methods to obtain the highest sensitivity in assessing variability in clinical activity.
Allocation was based on the direct assignment of the cost of the following services to the patient: laboratory, pharmacy, radiology, nuclear medicine, pathology, and prosthetics.
The information systems contain comprehensive data on human resources and their activity: storage, admissions planning, ambulatory, and emergency care, operating rooms, diagnostic and complementary tests, and inter-hospital consultation.
The main economic outcome was the total hospital cost of XDR P. aeruginosa and ESBL-K. pneumoniae episodes, including fixed costs, variable costs, and pharmacy costs. For the estimation of fixed costs, only episodes related to infection were compared. Fixed costs were derived from surgical procedures, hospitalization, and ICU-stay, and were allocated according to routine criteria: operation or intervention time or number of days of stay in the various hospital units. Variable costs included costs deriving from laboratory, radiology, pathology, prostheses, tests, and pharmacy. Additionally, costs derived from relapses of the infection were collected, including hospital readmissions.

Ethics
The study was approved by the Clinical Research Ethics Committee of the Parc de Salut Mar, Barcelona (register no. 2020/9321). Due to the observational nature of the study and retrospective analysis, the need for written informed consent was waived.

Results
In total, 1260 urine cultures positive for XDR P. aeruginosa or ESBL-K. pneumoniae during the study period were included in the preselection process (465 and 795, respectively). Based on the attempted sample size, 201 episodes that met the inclusion and exclusion criteria were included in the final analysis. Seven patients in the XDR P. aeruginosa cohort and eight in the ESBL-K. pneumoniae group experienced more than one episode. Two patients had infections caused by both microorganisms in different episodes. A total of 50/201 (24.8%) of the episodes were bloodstream infections.
The epidemiological and clinical characteristics of patients included in the study are shown in Table 1. Male sex was predominant in both cohorts, but significantly higher in the XDR P. aeruginosa group than in the ESBL-K. pneumoniae group (80 (79.2%) vs. 65 (65%); p = 0.025). No significant differences in age or Charlson comorbidity index were observed between the two groups. XDR P. aeruginosa episodes were more common in patients who had previously used urinary devices (p < 0.001), had urinary tract abnormalities (p < 0.001), or had nosocomial acquisition (p = 0.005). Patients with cUTI caused by XDR p. aeruginosa more frequently received inappropriate empirical therapy than those caused by ESBL-K. pneumoniae (80 (79.2%) vs. 44 (44%); p < 0.001) and also had longer delays to initiation of appropriate antibiotic therapy (54 (53.5%) vs. 39 (39%); p = 0.04).

Primary Outcome: Clinical Failure
Clinical failure on day 7 of treatment was 28.7% (29/101) and 8% (8/100) in the XDR P. aeruginosa and ESBL-K. pneumoniae groups, respectively (p < 0.001). Failure in the XDR P. aeruginosa group was due to the persistence or worsening of signs and/or symptoms (twenty-six episodes), death (two episodes), and one episode in which therapy had to be modified due to the side effects of antibiotics. Causes of failure in the ESBL-K. pneumoniae group were death (five episodes), persistence or worsening of signs and/or symptoms (two cases), and one case of treatment modification due to the side effects of antibiotics.
Clinical failure at the end of treatment was 19.8% (20/101) in the XDR P. aeruginosa cohort vs. 8% (8/100) in the ESBL-K. pneumoniae group (p = 0.016). Reasons for failure in the XDR P. aeruginosa group were persistence or worsening of signs and/or symptoms (nine episodes), need to modify therapy due to antibiotic side effects (four cases), isolation of a new strain of XDR P. aeruginosa resistant to antibiotic treatment (three episodes) and death (four patients). In all cases, the reason for switching antibiotic treatment following antibiotic side effects was nephrotoxicity. The reasons for failure in the ESBL-K. pneumoniae group were death (seven patients) and one patient with persistent or worsening of signs and/or symptoms.
The unadjusted and adjusted analysis of variables associated with clinical failure on day 7 and at EOT are shown in Tables 2 and 3, respectively. After adjusting for confounders in multivariate analysis, XDR P. aeruginosa cUTI was independently associated with clinical failure on day 7 of treatment (OR 4.34, 95% CI 1.71-11.04; p = 0.002) but not at the end of treatment (OR 2.31, 95% CI 0.83-6.44; p 0.108). Clinical failure at 7 days was also independently associated with quick SOFA (OR 1.97, 95% CI 1.11-3.51; p = 0.020). After adjusting for confounders, clinical failure at the end of treatment was independently associated with UTI classification (acute pyelonephritis) (OR 0.00, 95% CI −0.00), Charlson index (OR 1.20, 95% CI 1.01-1.43; p = 0.036) and quick SOFA (OR 2.19, 95% CI 1.15-4.18; p = 0.017), but not with XDR P. aeruginosa infection.

Adverse Events
Acute renal failure occurred only in the XDR P. aeruginosa infection group, in 27 patients (26.7%) who received treatment with colistin (n = 17), amikacin (n = 4) and amikacin + colistin (n = 6). The incidence of Clostridioides difficile infection was similar in both groups: 4% in XDR P. aeruginosa versus 5% in the ESBL-K. pneumoniae group. Neurological complications were seen only in two patients in the ESBL-K. pneumoniae group: one patient, on ertapenem treatment, had seizures; and one patient, on imipenem treatment, had an altered level of consciousness and myoclonus.

Discussion
The objective of our study was to compare clinical characteristics, outcomes and economic differences in cUTI caused by XDR P. aeruginosa and ESBL-K. pneumoniae. We found that patients with XDR P. aeruginosa cUTI had worse early clinical cures and more antibiotic side effects. However, we found no differences in clinical failure at end of treatment, mortality, or economic costs between the two groups. ESBL-K. pneumoniae cUTIs were associated with worse eradication rates and higher 90-day infection-related readmission. Although previous studies have associated healthcare-related UTI with increased resource consumption [37] and worse outcomes [16], few have compared the clinical and economic burden of infections caused by XDR P. aeruginosa versus other MDR-GNB pathogens [23,24].
In the present study, urinary tract infections caused by ESBL-K. pneumoniae were observed more frequently in patients with diabetes mellitus, in kidney transplant recipients, and with co-existing renal failure. In a review of the literature, different authors have shown comparable results [38][39][40]. In two retrospective case-control studies conducted to identify risk factors for ESBL-K. pneumoniae UTI, Espinar et al. found that diabetes mellitus (p < 0.007) and patients with delayed graft function (p = 0.001) were independent risk factors in non-hospitalized kidney transplant patients [38], while Lautenbach et al. found a borderline significant association between ESBL-Enterobacterales infection and diabetes (p = 0.07). In another case-control study comparing non-ESBL-K.pneumoniae vs ESBL-K.pneumoniae UTIs, chronic kidney disease was found to be more prevalent in the second group (p = 0.0296) [39]. The risk factors for XDR P. aeruginosa infection in our study were COPD and solid tumour malignancy, two conditions which are already known to be associated with P. aeruginosa infections. As an opportunistic pathogen, P. aeruginosa usually infects patients with immunodeficiency, malignancy, or chronic pulmonary disorder. In patients with chronic inflammatory airway diseases, such as COPD, cystic fibrosis, asthma, or bronchiectasis, P. aeruginosa produces chronic colonization of the lower respiratory tract and respiratory infections [41]. In these patients, P. aeruginosa frequently colonizes the gastrointestinal tract as well [42,43], making them more susceptible to cUTI by these pathogens than by other bacterium. Regarding patients with solid malignancy, their immunocompromised status and the use of immunosuppressive therapy promote P. aeruginosa infection in these patients [44].
With respect to clinical outcomes, our study suggests that XDR P. aeruginosa cUTIs have worse early clinical cure rates than ESBL-K. pneumoniae infections. It is also worth pointing out that a significant number of treatment failures in this group were due to antibiotic nephrotoxicity. It is known that P. aeruginosa can develop drug resistance during prolonged therapy as early as three days after the initiation of therapy with an antibiotic to which it is originally tested as sensitive, which would also further explain the difficulty of treatment and/or the poorer clinical outcome associated with this organism. [45]. These results, in agreement with our hypothesis, support the statement that infections caused by XDR P. aeruginosa are always a difficult-to-treat scenario with very limited treatment options that lead to worse clinical outcomes.
In terms of mortality, after adjusting for cofounders, we were unable to find differences between the two groups. Mortality rates for XDR P. aeruginosa infection (30-day mortality 9.9%, 90-day mortality, 24.8%) were lower than in previous reports: 33.6% for 30-day mortality in XDR P. aeruginosa infections [46] and around 30% for MDR/XDR P. aeruginosa episodes [20,22,47], although these studies included bacteraemic episodes and sources of infection other than the urinary tract. The differences in mortality therefore may be explained by the lower risk foci of infection of urinary tract infection, and low mortality rates [48]. On the other hand, 30-day mortality in our study in the ESBL-K. pneumoniae group (13%) was similar to other studies of mortality in UTIs caused by this microorganism [25,40]. In a cohort study by Richelsen et al. [25] evaluating the impact of ESBL production on community-onset infections due to Escherichia coli or K. pneumoniae, 30-day mortality was 13.8%. In another study by Larisa-Miftode et al. [40] comparing ESBL and non-ESBL-K. pneumoniae infections, mortality in the ESBL group was 17.6%.
Many different studies on ESBL-Enterobacterales and P. aeruginosa bloodstream infections have found that time to appropriate antibiotic treatment is an independent predictor of mortality [49][50][51][52]. However, for complicated urinary tract infections, there is no clear evidence. In our study, we found a considerable difference between the two groups in time to perform appropriate therapy (72 h delay: 53.5% in XDR P. aeruginosa and 39% in ESBL-K. pneumoniae). These differences were expected because there are fewer available treatment alternatives against XDR P. aeruginosa. Nevertheless, we did not find a statistically significant association with clinical failure or mortality. One explanation could be the lack of severity in the initial presentation of infection in most patients; the frequency of sepsis or septic shock was 31.7% and 32.7%, respectively. Some authors also argue that, in cases of P. aeruginosa infection, a 48-72 h delay of receipt of the appropriate antibiotics does not have a great impact on patient outcome, since mortality is mainly due to other factors, such as clinical presentation, source of infection or receipt of inappropriate definitive antibiotic therapy [53][54][55].
In recent decades, the increase in multi-drug resistance has an impact not only on clinical outcomes but also on economic costs. It is estimated that the cost of infections caused by antimicrobial-resistant organisms is higher, ranging from $6000-$30,000, than in infections caused by antimicrobial-susceptible organisms [56]. While previous studies have compared resource consumption between susceptible and resistant microorganisms, we set out to compare economic differences between two multidrug-resistant bacteria. In a review of the literature, several reports have associated XDR P. aeruginosa infection with higher resource consumption [57], although in our economic analysis we found no differences between the two microorganisms. Our results could be partly explained by the worse eradication and recurrence rates observed in the ESBL-K. pneumoniae group and thus, higher hospital readmission rates. While K. pneumoniae is a normal part of the intestinal flora in humans, hospital patients are more susceptible to colonization by this bacterium, which further complicates eradication [58]. Another possible explanation is that the patients with the highest prevalence of ESBL-K.pneumoniae cUTI were kidney transplant recipients and those with chronic renal failure. Both these comorbidities have been identified in previous studies as risk factors associated with UTI recurrence [59]. We suggest that some of the strategies to prevent cUTI recurrence in these patients should be based on the incorporation of antimicrobial stewardship programmes in healthcare facilities, promoting catheter restriction protocols to limit catheter use and appropriate discontinuation of catheters [60].
In the economic study, after adjustment for cofounders, bacteraemia and nosocomial acquisition were found to be associated with higher costs, which is explained by the greater severity of the infection and longer length of stay, respectively.
Our study has the inherent limitations of a retrospective, single-center study in a specific health system, which may influence the applicability of the results. In addition, our results may not be generalizable to other geographical areas with different epidemiology, especially regarding differences in the most prevalent high-risk clones of both XDR P. aeruginosa and ESBL-K. pneumoniae. Second, due to the retrospective nature of the study, we did not conduct molecular clonality studies of the strains, which would have been useful to confirm the presence of certain high-risk clones in our cohort, and to study resistance determinants. Third, since carbapenemase-producing K.pneumoniae may share more similarities than ESBL with XDR P. aeruginosa in terms of a spectrum of resistance, clonal widespread, and nosocomial/HCA acquisition, it could be of interest to explore the role of carbapenemase-producing K.pneumoniae as comparator group. However, since the prevalence of ESBL-K. pneumoniae is much higher in our geographical area [61,62], ESBL may reflect more accurately the burden of MDR Enterobacteriae in the study setting.
Four, our study was conducted at a time when the "new" antibiotics against MDR-GNB ceftolozane/tazobactam and ceftazidime/avibactam were not routinely tested. Furthermore, as mentioned before, mortality rates in urinary tract infections tend to be low compared with other sources of infection and the results are not therefore generalizable to other sources of infection. Finally, antibiotic dosing was not recorded. This information would have been useful to evaluate the role of appropriate antibiotic exposures and the impact on adverse effects.
As a strength, while several previous reports have analysed the clinical differences between resistant and susceptible P. aeruginosa infections, to our knowledge, this is the first report to compare the clinical impact of XDR P. aeruginosa with another difficult-to-treat bacteria (ESBL-K. pneumoniae), while also exploring the dimension of resource consumption.

Conclusions
Management of XDR P. aeruginosa infection is always challenging due to its intrinsic particularities and the limited treatment options available. In our study patients with XDR P. aeruginosa cUTI had worse early clinical cures and more antibiotic side effects, compared with ESBL-K.pneumoniae. We also found differences between the two groups in terms of appropriate time for empirical therapy, but there was no statistical association with clinical failure at the end of treatment or with mortality, probably explained by the low-risk foci of urinary tract infections. However, ESBL-K. pneumoniae cUTIs were associated with worse eradication rates, higher recurrence, and 90-day infection-related readmission. We did not find differences in either a clinical failure at the end of treatment, mortality, or economic costs between the two groups. Funding: This study was financed by a competitive grant from Fondo de Investigaciones Sanitarias at the Spanish government's National Institute of Health Research, (proyect FIS PI21/00509), funded by Instituto de Salud Carlos III (ISCIII) and co-funded by the European Union and by a research grant from the Spanish Society of Infectious Diseases and Clinical Microbiology (SEIMC). The work is also supported by the Spanish Network for Research in Infectious Diseases (REIPI) and Centro de Investigación Biomédica en Red Enfermedades Infecciosas (CIBERINFEC), Instituto de Salud Carlos III (Madrid, Spain), co-funded by the European Union.

Institutional Review Board Statement:
The study was conducted in accordance with the Declaration of Helsinki, and approved by the Clinical Research Ethics Committee of our institution (Clinical Research Ethics Committee of the Parc de Salut Mar, Barcelona; register no. 2020/9321). The need for informed written consent was waived because of the retrospective and observational nature of this study.

Data Availability Statement:
The data presented in this study is available in the article.