Risk factors for antimicrobial resistance in patients with Escherichia coli bacteraemia related to urinary tract infection

Summary Introduction NHS Lothian policy has recently changed to avoid first-line use of trimethoprim for uncomplicated urinary tract infections (UTI) in patients with risk factors for trimethoprim resistance, in line with national guidance. This study aimed to identify risk factors for antimicrobial resistance in Escherichia coli bacteraemia related to UTI. Methods A retrospective cohort study of 687 patients with E. coli bacteraemia related to UTI in NHS Lothian from 01/02/18 to 29/02/20 was undertaken. Demographics and comorbidities were collected from electronic patient records. Community prescribing and microbiology data were collected from the prescribing information system and Apex. Univariate and multivariate analysis was undertaken using RStudio to analyse trimethoprim, gentamicin and multi-drug resistance (MDR). Results Trimethoprim resistance was present in 282/687(41%) of blood culture isolates. MDR was present in 278/687(40.5%) isolates. Previous urinary trimethoprim resistant E. coli was a significant risk factor for both trimethoprim resistance (OR 9.44, 95%CI 5.83–15.9) and MDR (OR 4.81, 95%CI 3.17–7.43) on multivariate modelling. Trimethoprim prescription (OR 2.10, 95% CI 1.33–3.34) and the number of community antibiotic courses (OR 1.19, 95%CI 1.06–1.35) were additional risk factors for trimethoprim resistance. Multiple independent risk factors were also identified for trimethoprim resistance, MDR and gentamicin resistance. Discussion This study showed a high prevalence of trimethoprim resistance and MDR in patients with E. coli bacteraemia related to UTI. This supports the withdrawal of trimethoprim from first-line treatment of UTIs in patients with risk factors for trimethoprim resistance. It has also identified risk factors for MDR in E. coli bacteraemia.


Introduction
Antimicrobial resistance (AMR) is a significant and increasing problem worldwide. Urinary tract infections (UTI)s are a common cause of morbidity and mortality with severe complications including bacteraemia and sepsis. Escherichia coli is the most common organism isolated from urinary samples in Scotland, England and Europe [1e3]. E. coli bacteraemia is strongly associated with UTI and accounts for the majority of Gramnegative bacteraemias in Scotland [4]. AMR is highly prevalent in urinary E. coli and E. coli bacteraemia in Scotland [4,5]. E. coli bacteraemia is associated with 30-day all-cause mortality from [8][9][10][11][12][13][14][15][16][17][18].2% in hospital inpatients [6e8] and associated AMR leads to less effective management [7,9e11].
E. coli was isolated from 115,844 urine samples and 4,206 blood cultures across Scotland in 2020 [1]. The incidence of E. coli bacteraemia was 77 per 100,000 population [1]. This has been broadly stable over the last five years [1,4]. The high incidence of trimethoprim resistance in urinary E. coli led to NICE recommending avoidance in patients at risk of resistance. Risk factors identified included hospital inpatients, nursing or care home residents, over 65-year-olds, trimethoprim prescription in the previous three months, or trimethoprim resistant organisms in the previous three urine cultures, over the previous 12 months [12]. NHS Lothian guidelines for community antibiotic prescribing changed to reflect this after the data for this study was collected. In 2019, the (UK) Government set the target of a 50% reduction in Gram-negative bacteraemias by 2023/2024. [4] This study aimed to examine the risk factors for AMR and MDR in E. coli bacteraemia related to UTI to allow the review and optimisation of local empirical antimicrobial prescribing guidelines and subsequently reduce the number of E. coli bacteraemias related to UTI.

Methods
This was a retrospective cohort study of all patients with E. coli bacteraemia associated with UTI in blood cultures across NHS Lothian from 01/02/2018 to 29/02/2020. In total, 710 E. coli bacteraemia isolates related to UTI were identified.
Repeat isolates and paediatric cases were removed leaving 687 cases. Clinical data, demographics and comorbidities were drawn from Trak (online patient records) and microbiology data from Apex (laboratory systems). Demographic data included age, gender and nursing home residence. Comorbidities included diabetes mellitus, previous urological surgery and long-term catheterisation. Urinary E. coli isolates in the previous 12 months were analysed. Community prescribing data was collected over 6 months before the bacteraemia from the prescribing information system (PIS).
Data manipulation, univariate and multivariate analyses were carried out in RStudio version 4.1.0. Multivariate models were evaluated using the Akaike information criterion (AIC) with a backwards stepwise method using the 'MASS' package [13]. Collinearity between risk factors was allowed for.
All isolates received standard sensitivity testing using VITEKÒ 2 automated broth microdilution cards N381 (blood cultures) and N382(urines) andcategorised using EUCASTbreakpointsv8.0,v8.1 and v9.0 [14]. Whilst there is no standard bacteraemia breakpoint for trimethoprim resistance, this was extrapolated from urinary breakpoints as a minimum inhibitory concentration of >4mg/L. Gentamicin is the first-line treatment of upper UTI and sepsis secondary to urinary source within NHS Lothian guidelines, so was also analysed [15].Thereisnogloballyaccepteddefinitionfor MDR and extensively drug resistant (XDR) Enterobacteriaceae. The ECDCdefinition [16]isnotsuitedforroutineclinicaluseasitutilises 17 different antimicrobial categories and contains agents, including entire classes, which are not routinely tested for in Scotland. For this reason, the Canadian system for describing MDR and XDR Enterobacteriaceae, described by German et al. (2018), currently under proposal for use in NHS Lothian, was used to categorize MDR and XDR isolates [17]. Isolates were described as MDR if they were resistant to three or four of the six groups of antimicrobialslistedbelow.XDRwasdefinedasgreaterthanfourofthe six groups. Antimicrobial sensitivities were divided into the followinggroups:tobramycin/gentamicin,piperacillin-tazobactam, imipenem/meropenem, cefotaxime/ceftriaxone/ceftazidime, ciprofloxacin and trimethoprim-sulfamethoxazole [17]. In our analysis trimethoprim was used to indicate trimethoprimsulfamethoxazole susceptibility. Intermediate resistance, as per EUCAST breakpoints v12.0, was included with resistant isolates as these antibiotics would be avoided in practice [14].

Results
Demographics 687 patients with UTI related E. coli bacteraemia were examined with a mean age of 71.3 years. There were 407 (59.2%) female patients. Nursing home residents made up 47/ 687 (6.84%). Insulin-dependent diabetes was present in 50/687 (7.28%) patients. Long term catheters were present in 162/687 (23.58%) patients. Previous urological surgery had occurred in 94/687 (13.68%) patients. Prophylactic antibiotic prescription in the community occurred in 33/687 (4.80%) patients. The proportion of isolates with AMR to commonly used antibiotics was measured across the cohort ( Figure 1).
Multivariate analysis showed that previous trimethoprim resistant urinary E. coli (OR 4.81, 95%CI 3.17e7.43) and the number of community antibiotic courses were significant risk factors for MDR (Table II).
Only 8/687 cultures were extremely drug-resistant (XDR) therefore no additional analyses were undertaken.

Discussion
The demographics and resistance profile of the cohort were concordant with Scottish, English and European surveillance data [1e3]. This study showed that previous urinary trimethoprim resistant E. coli was the most significant risk factor for trimethoprim resistance and MDR in E. coli bacteraemia. Trimethoprim resistance was also associated with trimethoprim prescription and the number of antibiotic courses prescribed. MDR was associated with the number of community antibiotic courses prescribed. This agrees with the multiple studies demonstrating community antibiotic prescribing is linked to AMR in E. coli urinary and blood isolates [5,18,19]. This study   [5]. However, the present study showed that nitrofurantoin prescription was negatively associated with trimethoprim resistance. This relationship requires further exploration, particularly as nitrofurantoin use may increase with the change in local guidelines.
In the present study, clinical factors including age, diabetes mellitus, long-term catheterisation, and previous urological surgery were assessed. Age and long-term catheterisation were independent risk factors for gentamicin resistance. Prophylactic antibiotic therapy in the community was independently linked to both trimethoprim resistance and MDR. In a recent study, Aliabadi et al. (2021) analysed risk factors for AMR in E. coli bacteraemia in 175,147 patients from English national surveillance data. Increasing age and regional deprivation were associated with AMR in community-acquired E. coli bacteraemia [20].
In the present study, univariate analysis revealed a correlation between trimethoprim resistance in urinary E. coli and gentamicin resistance in E. coli bacteraemia. This effect, alongside the failure of multivariate modelling for gentamicin resistance, may be due to the small sample size. However, this link should be explored further to allow the optimisation of community prescribing.
The strengths of this study include the analysis of blood culture isolates, guaranteeing the significance of the infection. Some urinary isolates may reflect asymptomatic bacteriuria. However, it has been suggested that AMR bacteria may be more likely to progress to bacteraemia [10]. This study analysed resistance to trimethoprim, gentamicin and MDR, rather than a single antimicrobial.
The limitations of this study include a lack of inpatient prescribing data. Patients may have received antibiotics as an inpatient prior to the development of the bacteraemia. Temporal data from the community prescribing was not collected therefore the length of time from prescription to bacteraemia could not be analysed.

Conclusion
Trimethoprim resistance and MDR are prevalent in E. coli bacteraemia associated with UTI. The most significant risk factor for trimethoprim resistance and MDR was previous urinary trimethoprim resistant E. coli. Community trimethoprim prescription and cumulative community antibiotic courses were also associated with trimethoprim resistance. Many factors, particularly community antibiotic prescribing, were independently linked to AMR and MDR. This study supports the move away from first-line trimethoprim use for UTIs in patients with risk factors for resistance. Since the data collection period, NHS Lothian has moved away from first-line trimethoprim use in patients at risk of resistance. This data should be considered to further optimise empirical prescribing guidelines in the hope it may lead to a reduction in E. coli bacteraemias.