Intestinal colonisation with extended-spectrum beta-lactamase-producing Enterobacterales and the impact of selected factors on intestinal colonisation duration: a prospective cohort study

Abstract

Introduction

We conducted a prospective cohort study of patients with newly acquired extended-spectrum beta-lactamase-producing Enterobacterales (ESBLPE) intestinal colonisation (termed “ESBLPE colonisation”) in two public hospitals between January 2013 and October 2016. We evaluated the duration of ESBLPE colonisation and the impact of selected factors on this duration.

Methods

Patient data and faeces samples were collected on enrolment and three monthly for 2 years. Standard laboratory methods were used for ESBLPE identification and susceptibility testing. Whole-genome sequencing of isolates was performed determining sequence type, plasmid replicons, antimicrobial resistance genes, virulence factor genes and phylogenetic relationships.

Results

The median duration of ESBLPE colonisation for the 102 patients was 544 days (range 77–730 days, interquartile range 220–730 days). After one year and at the end of follow up, 61% and 23.4% of patients respectively remained ESBLPE colonised. The same strain of ESBLPE, colonised 70.2% (33/47) patients colonised with Escherichia coli and 70.7% (29/42) patients colonised with Klebsiella pneumoniae. Prolonged ESBLPE colonisation was associated with age ≥ 75 years (HR 0.46, 95% CI 0.23–0.90), an increased intensity of hospitalisation (P = 0.0036), E. coli phylogroup B2 (P = 0.0142), E. coli harbouring bla_CTX−M−15 (P < 0.0001), colonisation with more than one species of ESBLPE (P = 0.0023) and colonisation with more than one E. coli Sequence Type (P = 0.0042).

Conclusions

It is likely that combinations of multiple patient and bacterial factors influence the duration of ESBLPE colonisation. This prolonged duration of ESBLPE colonisation highlights the importance of infection control and public health measures to decrease the acquisition and spread of ESBLPE.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13756-025-01685-5.

Introduction

Extended-spectrum beta-lactamase (ESBL)-producing Enterobacterales (ESBLPE) emerged in the 1980s in Europe. Initially, ESBLPE were observed mainly in hospitals and were most commonly Klebsiella spp., producing SHV or TEM enzymes [1]. In the following decades ESBLPE spread throughout the world and, in the mid-2000s, into the community, with CTX-M-producing Escherichia coli joining Klebsiella spp. as a major host [2]. A recent systematic review and meta-analysis reported that global human intestinal colonisation with ESBL-producing E. coli is increasing in both healthcare and community settings [3]; the global colonisation rate in healthcare settings increased 3 to 4-fold from 7% in 2001-05 to 25.7% in 2016-20, and in community settings increased 10-fold from 2.6 to 26.4% during the same period. At the same time, CTX-M enzymes have replaced SHV and TEM enzymes as the globally predominant ESBLs [4]. ESBLPE are often multiresistant, significantly reducing antibiotic choice for the treatment of infections due to these organisms and challenging infection control strategies [5].

Knowledge about the duration of ESBLPE intestinal colonisation (subsequently termed ‘ESBLPE colonisation’) and the factors influencing its duration are important for the implementation of adequate infection control measures and improved patient management.

Prospective studies that have followed adult patients for at least one year, reported that ESBLPE colonisation was still present in 22–44% of patients at 1 year, 19–31% at 2 years and 15% at 3 years [6–8]. A Swedish study [9] found that five of 42 patients still carried ESBL-producing E. coli after 41–59 months duration, while a ten-year retrospective study [10] found that colonisation with the same strain persisted for a median of 231 days (IQR 106–535) for K. pneumoniae and 332 days (IQR 170–780) for E. coli.

Various risk factors are associated with prolonged ESBLPE colonisation, including E. coli isolates of phylogroup B2 [6, 7] or D [6], ESBL-CTX-M-group 9 [7], immobility [8] and immunosuppression [10]. Some investigators have not found any risk factors associated with prolonged colonisation [9].

We conducted a prospective cohort study of patients with newly acquired ESBLPE colonisation, aiming to evaluate the duration of colonisation and the impact of selected factors on this duration.

Methods

Study setting

The study was conducted from January 2013 to October 2016 in two large public hospitals in Auckland, New Zealand. Each hospital has a catchment population of greater than 500,000 people.

Patient data and specimen collection

Patients were recruited from January 2013 until October 2016. All patients admitted to the two hospitals who were considered to be at risk of being colonised with a multidrug-resistant organism (MRO) were screened for ESBLPE colonisation as part of a regular screening programme. At the time of study enrolment, the criteria for MRO screening, was any patient who had been admitted to any New Zealand or overseas hospital in the previous six months. Adult inpatients who were 16 years or older and found to be newly colonised with an ESBLPE were asked to participate in the study. A patient with a rectal swab or faeces sample yielding an ESBLPE was considered to be newly colonised if there was no record of previous ESBLPE colonisation or infection. Exclusion criteria were as follows: age less than 16 years, not able or willing to give informed consent or comply with study protocol and unable to communicate in English.

Patient data were collected on enrolment and included information on demographics, history of hospitalisation and antibiotic use in the previous year, history of international travel in the previous two years, presence of active medical conditions, presence of chronic skin conditions, a current episode of ESBLPE infection, antimicrobial treatment, immunosuppressive treatment, and the presence of an indwelling device on discharge from hospital. Data were collected directly from patients and their medical record.

Patients were then followed prospectively for 2 years and were asked to submit a faeces sample and to complete a questionnaire every 3 months. The questionnaire sought information from the previous three-month period including place of residence, visits to a general practitioner, hospitalisations, occurrences of ESBLPE infection, presence of a chronic skin condition, antibiotic use, presence of a urinary or dialysis catheter and episodes of international travel. Antibiotic use during hospitalisations and in the community was obtained from the patients’ medical and regional electronic records respectively.

The analysis of ESBLPE colonisation was performed on patients who were followed up for ≥ 12 months and had at least one positive sample or two negative samples in the second year of follow up (Fig. 1).

Fig. 1.

Number of patients included in the various components of the analysis

Definitions

• Study enrolment date was the collection date of the initial sample yielding an ESBLPE.

• Duration of follow up was the number of days from the study enrolment date to the collection date of the last sample.

• Duration of colonisation was the number of days from the study enrolment date until the collection date of the first negative sample after the last ESBLPE positive sample for the patients who no longer tested positive for an ESBLPE or from the study enrolment date until the collection date of the last ESBLPE positive sample for the patients whose last sample remained ESBLPE positive.

• Prolonged ESBLPE colonisation was the presence of ESBLPE positive samples for ≥ 12 months after the study enrolment date.

• Long colonisation group was the group of patients who had one or more ESBLPE positive samples present for ≥ 12 months after the study enrolment date.

• Short colonisation group was the group of patients who had ESBLPE positive samples present for < 12 months from the enrolment date, with at least two negative samples and no ESBLPE positive samples ≥ 12 months after the study enrolment date.

• Intensity of antibiotic use (ABI) was the duration of antibiotic use (in days) divided by the duration of the follow-up (in days) from the study enrolment date to the date of the first negative sample after the last ESBLPE positive sample for the patients who no longer tested positive for an ESBLPE or from the study enrolment date until the date of the last ESBLPE positive sample for the patients who remained ESBLPE positive.

• Intensity of duration of hospitalisation (DHI) was the duration of hospitalisation (in days) divided by the duration of the follow-up (in days) from the study enrolment date to the date of the first negative sample after the last ESBLPE positive sample for the patients who no longer tested positive for an ESBLPE or from the study enrolment date until the date of the last ESBLPE positive sample for the patients who remained ESBLPE positive.

Laboratory methods

Detection, identification and whole-genome sequencing of ESBLPE isolates

Initial samples were collected using a rectal swab. Follow up faeces samples were collected in a sterile container. Selective chromogenic plates (Brilliance ESBL Agar (Oxoid, UK) and ChromID ESBL Agar (BioMerieux, Marcy-l’Etoile, France)) were directly inoculated without previous broth enrichment. Colonies resembling Enterobacterales were identified to species level using MALDI-TOF mass spectrometry, Biotyper 3.0 software (Bruker Daltonics, Bremen, Germany) and Vitek MS (BioMerieux, Marcy-l’Etoile, France), respectively. ESBL production was confirmed using Clinical and Laboratory Standards Institute’s combination disc diffusion method [11] or by testing a suspension of the isolate for CTX-M group 1 or 9 ESBL genes (AusDiagnostics now R-Biopharm AG multi-plex panels).

Faeces samples that were negative for an ESBLPE on direct culture were inoculated into a MacConkey broth (Fort Richard Laboratories). After overnight incubation, a PCR assay (AusDiagnostic now R-Biopharm AG multi-plex panels) was used to detect group 1 and 9 CTX-M ESBLs in the enrichment broth. Media with detectable ESBL genes were then subcultured on ChromID ESBL Agar (BioMerieux, Marcy-l’Etoile, France) in an attempt to obtain the ESBLPE isolate. A sweep of ESBLPE was stocked in skimmed milk at -70° C for further analysis. If different morphologies were observed, these were also stocked.

Whole-genome sequencing (WGS) of isolates was carried out at the Australian Centre for Ecogenomics Sequencing Facility. Genomic libraries were prepared using Nextera XT library prep (Illumina) and sequenced on a Nextseq 500 (Illumina) using a 2 × 150 bp High Output V2 kit (isolates MS14620-MS14934, n = 263, Supplementary Table 1). Samples showing evidence of mixed-species content in initial quality checks were sub-cultured, and colonies with distinct morphologies stocked as separate isolates. Genomic DNA extraction (DSP DNA Mini kit on the QIASymphony SP [Qiagen]) and WGS (Nextera XT DNA preparation kit Illumina] of these additional isolates (and those with low sequence yield initially) was carried out at the Queensland Forensic Scientific Services sequencing facility (isolates MS14948-MS15016, n = 57, Supplementary Table 1).

Bioinformatics workflow and methods

Genomic analysis was performed using in-house microbial genomic analysis pipeline, SnapperRocks (https://github.com/FordeGenomics/SnapperRocks). In brief, the quality of reads was checked using FastQC v0.11.6 [12] and low-quality reads and adapters were trimmed using Trimmomatic v0.36 [13]. Contamination detection was performed using CheckM2 v1.02 [14] and taxonomic classification was performed using Kraken v2.0.8 [15]. De novo assembly was performed using SPAdes v3.14.0 [16]. Low-quality contigs (contigs length < 100 bp; coverage < 20X) were removed. Assemblies with CheckM2 contamination > 10% and unusual large total length or has too many short contigs (> 1000 contigs) were excluded. Antimicrobial resistance genes (ARGs) typing and multi-locus sequence typing (MLST) were performed in silico using AMRFinderPlus v3.10.24 [17] and mlst v.2.19 with scheme retrieved from PubMLST database [18, 19]. Plasmid replicon and virulence factors typing was performed using Abricate v1.01 (https://github.com/tseemann/abricate) against Plasmidfinder [20] and VFDB [21] database, respectively. The BLAST nucleotide identity was used for each resistance, replicon and virulence gene. With the exception of bla_CTX−M (86%) all other resistance genes used in the phylogenetic tree figures (Figs. 2 and 3 and Suppl. Figures 4–8) had a BLAST nucleotide percentage identity of 100% and minimum coverage of 100%. In the replicon and virulence factor gene datasets the minimum coverage cutoff and the minimum BLAST nucleotide identity cutoff was 90%. In nearly all cases the BLAST nucleotide identity was 98–100% identity.

Fig. 2.

Mid-point rooted maximum likelihood phylogenetic tree of E. coli ST131 based on 9275 core genome SNPs against a lineage reference genome (Genbank Accession: HG941718). Isolates from same patient are highlighted by same colour. Heatmap columns represent presence (grey) and absence (white) of identified beta-lactamase genes and plasmid replicon types. Scale represents nucleotide substitutions per site

Fig. 3.

Mid-point rooted maximum likelihood phylogenetic tree of K. pneumoniae ST25 based on 249 core genome SNPs against a species reference genome (Genbank Accession: CP003200). Isolates from same patient are highlighted by same colour. Heatmap columns represent presence (grey) and absence (white) of identified beta-lactamase genes and plasmid replicon types. Scale represents nucleotide substitutions per site

Whole genome species level tree was performed using Mashtree v1.2.0 [22] with bootstrap (--reps 1000). Sequence type level trees were performed using SNPDragon v1 [23]. Briefly, SNPDragon performed read mapping using BWA, variant calling using FreeBayes, and variant filtering (removal of sequential SNPs based on sliding window approach) against a reference genome to extract core genome SNPs alignment and pairwise distance matrix. The SNPs alignment was used to construct phylogenetic tree using IQ-TREE 2 v2.1.2 [24] with GTR Gamma Invariable site model (GTR + I + G) and 1000 bootstrap. Visualisation of the trees was performed using ggtree v3.10 R package [25, 26]. For SNP distance visualisation between close isolates from the same patient (same ST), a core SNPs alignment was generated using ParSNP [27] against publicly available same ST reference genome retrieved from Enterobase and NCBI genome (accessed on 4 April 2024). SNP distances were interactively inspected and visualised using GraphSNP [28].

Statistical methods

Statistical analysis was performed using SAS^® v9.4 software (SAS Institute Inc., Cary, North Carolina, USA). Frequency and proportion (%) were used to describe categorical variables. For continuous variables, mean, median, range and interquartile range were reported when appropriate. The Chi-square test was used when appropriate for comparisons of the categorical variables. Fisher’s exact test was used in situations where the Chi-square test was not appropriate for the above comparisons. The Kaplan-Meier Survival Plot was used to show the overall change of the ESBLPE colonisation rate, with median and 95% confidence intervals. Multiple Cox regression model was used to study the factors contributing to prolonged ESBLPE colonisation. A P-value of < 0.05 (two sided) was considered to be statistically significant.

Results

One hundred and fifty-six patients met the inclusion criteria and were enrolled. Thirty patients did not complete any follow up requirements after enrolment and so were subsequently excluded from the study (Fig. 1). The remaining 126 patients, who had at least one sample collected during the follow-up period, form the evaluable study cohort. Their characteristics are presented in Table 1. Comparing the baseline characteristics of the evaluable study cohort (n = 126) and the excluded patients (n = 30), showed differences in self-reported ethnicity with an increased proportion of Pacific peoples and a decreased proportion of Europeans in the group that did not receive follow up (P = 0.004) and in the proportion of patients who reported a previous contact with an ESBLPE colonised person with an increased proportion in the group that did not receive follow up (P = 0.024).

Table 1.

Baseline characteristics of the evaluable study cohort (n = 126)

Baseline variables		Number	%
Age (years)	16–64	51	40.5
	65–74	32	25.4
	75+	43	34.1
Sex	Female	55	43.7
	Male	71	56.4
Ethnicity	European	106	84.1
	Māori	9	7.1
	Pacific Peoples	2	1.6
	Asian	6	4.8
	MELAA/Other	3	2.4
Residence	Home	120	95.2
	Long-term care facility	6	4.8
Family member known to be colonised with an ESBLPE	No or don’t know	123	97.6
	Yes	3	2.4
Known contact with ESBLPE positive person	No or don’t know	121	96.0
	Yes	5	4.0
Overseas travel in the two years before the enrolment	No	64	50.8
	Yes	62	49.2
Hospital admission(s) in the 12 months prior to enrolment	No	47	37.3
	Yes	79	62.7
Antibiotic use in the year before the enrolment	No	14	11.1
	Yes	112	88.9
Active medical problems, including surgery	No	74	58.7
	Yes	52	41.3
Chronic skin conditions	No	112	88.9
	Yes	14	11.1
ESBLPE infection on enrolment	No	104	82.5
	Yes	22	17.5
Indwelling devices	No	107	84.9
	Yes	19	15.1
Antibiotic use from specimen collection until hospital discharge	No	34	27.0
	Yes	92	73.0
Immunosuppressive treatment from specimen collection until hospital discharge	No	115	91.3
	Yes	11	8.7

MELAA: Middle Eastern/Latin American/African, ESBLPE: extended-spectrum beta-lactamase-producing Enterobacterales

Of the evaluable study cohort, there were 18 patients who received less than one year of follow up and six patients who had only one negative sample and no positive samples after one year of follow up. These 24 patients were not included in the ESBLPE colonisation analysis (Fig. 1). There was no difference in baseline characteristics between the 102 patients included in and the 24 patients not included in the ESBLPE colonisation analysis. Of the 102 patients included in the ESPLPE colonisation analysis, 49 met the definition for the short colonisation group and 53 for the long colonisation group.

Seven hundred and eighty-seven samples were collected from the 126 patients (median 7, range 2–10 per patient). Three hundred and ninety-three (49.9%) samples yielded an ESBLPE; 367 (46.6%) from direct culture, 14 (1.8%) from enrichment culture only and 12 (1.5%) had ESBL genes detected by PCR only.

There were 439 isolates cultured; 425 (96.8%) isolates from direct culture and 14 (3.2%) isolates from enrichment culture. E. coli was the most common isolate [266/439 (60.6%)], followed by K. pneumoniae [160/439 (36.5%)]. For some of the remaining 13 isolates, the genotypic identification differed from the phenotypic identification, and the former was accepted as the final identification: Klebsiella michiganensis (3 isolates, phenotypic identification: Klebsiella oxytoca), Klebsiella aerogenes (2 isolates), K. oxytoca (1 isolate), Klebsiella variicola (1 isolate, phenotypic identification: K. pneumoniae), Enterobacter cloacae (2 isolates), Enterobacter hormaechei (2 isolates, phenotypic identification: E. cloacae), Citrobacter portucalensis (1 isolate) and Proteus mirabilis (1 isolate) [29, 30, 31].

Whole-genome sequencing (WGS) was available for 320 isolates obtained from 89 of the 126 patients. The isolates were E. coli (178 isolates), K. pneumoniae (131), K. michiganensis (3), K. aerogenes (2), K. oxytoca (1), K. variicola (1), E. hormaechei (2), C. portucalensis (1) and P. mirabilis (1).

The most prevalent ESBL gene was bla_CTX−M−15, which was identified in 217/320 (67.8%) isolates. Most (125, 95.4%) K. pneumoniae and less than a half (82, 46.1%) of the E. coli contained the gene. The second most prevalent ESBL gene was bla_CTX−M−14, identified in 64 (35.9%) E. coli and only two (1.5%) K. pneumoniae. There were only five isolates with no bla_CTX−M gene. The full extent of ESBL and other resistance genes is shown in Supplementary Fig. 1.

The 178 E. coli isolates encompassed 27 sequence types (ST), with five predominant STs (STs with ≥ 10 isolates): ST131 (34.8%), ST69 (15.2%), ST12 (9%), ST357 (8.4%) and ST38 (7.9%). The 131 K. pneumoniae isolates encompassed 16 STs, with ST25 (59.5%) and ST48 (13%) being the two predominant STs. The full list of STs identified in all species is provided in Supplementary Table 1.

Phylogroup B2 was the most common among E. coli (103/178, 57.9%), followed by phylogroup D (44, 24.7%), A (13, 7.3%), B1 (11, 6.2%) and F (7, 3.9%). Extraintestinal Pathogenic E. coli (ExPEC) lineages [32] were predominant in phylogroup B2 (84 isolates, 81.6%). ST131 was the most common (62 isolates, 60.2%), followed by ST12 (16, 15.5%), ST95 (4, 3.9%) and emerging fluoroquinolone-resistant lineage ST1193 (2, 1.9%) [33]. Thirty-one ST131 isolates (50%) harboured bla_CTX−M−14, while 19 (30.7%) isolates contained bla_CTX−M−15.

In E. coli, the most abundant virulence factor-encoding genes (VFs), detected in at least 100 E. coli isolates (≥ 100/178, ≥ 56.2%) were VFs for the synthesis of adhesins (fimABCDEFGHI, fdeC, csgBDFG, yagKVWXYZ/ecpREDCBA, ykgK/ecpR, papBX); factors involved in iron capture and metabolism such as siderophores enterobactin (entABCDEFS, fepABCDG, fes) and yersiniabactin (fyuA, ybtAEPQSTUX, irp1, irp2), chuASTUVWXY; capsule (kpsDM), arylsulfatase (asIA), general secretory pathway (gsp) and outer membrane protein (ompA) secretion.

In K. pneumoniae, the most abundant VFs (≥ 100/131 isolates, ≥ 76.3%) were VFs for the synthesis of adhesins (yagVWXYZ/ecpEDCBA, ykgK/ecpR), siderophores enterobactin (entAB, fepC) and yersiniabactin (fyuA, irp1, irp2, ybtAEPQSTUX), and outer membrane protein (ompA).

All VFs detected in ≥ 20 isolates and grouped by species are shown in Supplementary Fig. 2.

Incompatibility group F (IncF) plasmids were most commonly detected. In E. coli, IncFIB (AP001918) and IncFIA were the most common plasmid replicons, detected in 119/178 (66.9%) and 97/178 (54.5%) isolates, respectively. In K. pneumoniae, IncFII (pKP91) and IncFIB (Kpn3) were the most common replicons, detected in 131/131 (100%) and 105/131 (80.2%) of isolates, respectively. All detected plasmid replicons are grouped by species and shown in Supplementary Fig. 3.

Phylogenetic analysis revealed that most isolates of the same ST from the same patient clustered together. This indicated colonisation with the same strain over the course of the study for 70.2% (33/47) of patients colonised with E. coli and for 70.7% (29/41) patients colonised with K. pneumoniae, who had their isolates sequenced. This was supported by the colonisation with identical or almost identical resistance gene profiles and plasmid types (Figs. 2 and 3, Supplementary Fig. 4–8). The SNP distance between the isolates considered to be the same strain varied by species and ST (e.g. median of 1–14 within-cluster SNP differences in E. coli ST131). Occasional isolates of the same ST did not cluster with other isolates obtained from the same patient. One such example is NSH32, where two E. coli ST131 isolates collected ~ 10 months apart differ by 4608 SNP differences, consistent with independent acquisition of two unrelated ST131 strains.

ESBLPE colonisation analysis

The median follow-up for the 102 patients was 728 days (range 383–731 days; interquartile range 716–730 days). Consistent colonisation with one same bacterial species producing an ESBL (73/102, 71.6%) was more common than colonisation with two or more ESBL-producing bacterial species. ESBLPE were not consistently detected in the follow up samples taken during the period of sample positivity. There were instances where one or more negative samples were followed by a positive sample of the same or a different species.

The median duration of colonisation for the 102 patients was 544 days (range 77–730 days, interquartile range 220–730 days). After one year and at the end of follow up, 61% and 23.4% of patients respectively remained ESBLPE colonised (Fig. 4).

Fig. 4.

Proportion of patients remaining ESBLPE colonised during the follow up period

Risk factors associated with prolonged ESBLPE colonisation

Univariate analysis of patients in the short colonisation group compared with patients in the long colonisation group found no baseline characteristics associated with prolonged ESBLPE colonisation (Supplementary Table 2). However, univariate analysis of follow up data showed DHI was associated with prolonged ESBLPE colonisation (P = 0.0036) (Table 2).

Table 2.

Univariate analysis of potential risk factors associated with prolonged ESBLPE colonisation

Follow up variables		Short colonisation		Long colonisation		P value
		Number	%	Number	%
Residence	Home	45	91.84	44	83.02	0.1821$
	Long-term care facility	4	8.16	9	16.98
Intensity of duration of hospitalisation	Median (IQR#) Range	0% (1.5%) 0-23.9%		1.4% (4.1%) 0-28.9%		0.0036*
Intensity of antibiotic use	Median (IQR#) Range	7.4% (23.1%) 0-84.9%		7.9% (11.4%) 0-90.1%		0.6778*
Seen General Practitioner	No	3	6.12	1	1.89	0.3485**
	Yes	46	93.88	52	98.11
Urinary catheter placed	No	36	73.47	36	67.92	0.5392$
	Yes	13	26.53	17	32.08
Dialysis catheter placed	No	48	97.96	52	98.11	1.0000**
	Yes	1	2.04	1	1.89
Developed chronic ulcers	No	42	85.71	41	77.36	0.2788$
	Yes	7	14.29	12	22.64
Developed eczema/dermatitis	No	36	73.47	33	62.26	0.2268$
	Yes	13	26.53	20	37.74
Developed ESBLPE infection	No	43	87.76	45	84.91	0.6761$
	Yes	6	12.24	8	15.09
Overseas travel during follow-up	No	32	65.31	37	69.81	0.6270$
	Yes	17	34.69	16	30.19

ESBLPE: extended-spectrum beta-lactamase-producing Enterobacterales, $ Chi-square test, # interquartile range (Q3-Q1), * Wilcoxon rank sum test, ** Fisher's exact test

In order to control for the potential biases introduced by the duration of antibiotic use or the duration of hospital admission(s), ABI, DHI and age group were used in a multiple Cox regression model analysis. Higher DHI, ≥ 2% vs. no admissions (hazard ratio 0.42, 95%CI 0.21–0.82) and < 2% vs. no admissions (hazard ratio 0.36, 95%CI 0.19–0.70) as well as being 75 years of age or older (hazard ratio 0.46, 95%CI 0.23–0.90) were found to be associated with prolonged ESBLPE colonisation. ABI was not associated with prolonged ESBLPE colonisation (Table 3).

Table 3.

Multiple Cox regression model analysis of potential risk factors associated with prolonged ESBLPE colonisation

Risk factor	Comparison	P value	Hazard ratio	95% Hazard ratio confidence limits
Intensity of antibiotic use (reference: lower quartile)	Upper quartile	0.6293	0.816	0.357	1.864
	Median to upper quartile	0.5851	0.781	0.321	1.897
	Lower quartile to median	0.3410	0.689	0.321	1.482
Age group (reference: 16–64 years)	75 + years	0.0239	0.459	0.234	0.902
	65–74 years	0.2538	0.668	0.333	1.336
Intensity of duration of hospitalisation (reference: no admissions)	≥ 2%	0.0110	0.418	0.214	0.819
	< 2%	0.0027	0.363	0.187	0.703

ESBLPE: extended-spectrum beta-lactamase-producing Enterobacterales

In order to assess whether patient comorbidity was associated with prolonged ESBLPE colonisation, we included the Charlson comorbidity index score without inclusion of the age [34] in a revised Cox regression analysis. This showed that the Charlson comorbidity index score independent of the patient’s age did not contribute to the duration of ESBLPE colonisation (hazard ratio 1.02, 95% CI 0.84–1.23).

Most patients were colonised with either E. coli and/or K. pneumoniae (100/102, 98.4%). Sixty-six patients had the majority of their E. coli and K. pneumoniae isolates sequenced.

Potential bacterial characteristics contributing to prolonged ESBLPE colonisation were colonisation with more than one ESBL-producing bacterial species (P = 0.0023), E. coli harbouring bla_CTX−M−15 (P < 0.0001), E. coli phylogroup B2 (P = 0.0142) and colonisation with more than one E. coli ST (P = 0.0042) (Table 4).

Table 4.

Potential bacteria-related characteristics associated with prolonged ESBLPE colonisation

Bacteria related characteristics		Short colonisation group	Long colonisation group	P value
Number of patients consistently colonised with one same bacterial species		42/49 (85.7%)	31/53 (58.5%)	0.0023 $
Number of patients with E. coli phylogroups	A	1/18 (5.6%)	5/29 (17.2%)	0.3839*
	B1	2/18 (11.1%)	4/29 (13.8%)	1.0000*
	B2	8/18 (44.4%)	23/29 (79.3%)	0.0142 $
	D	5/18 (27.8%)	10/29 (34.5%)	0.6317$
	F	3/18 (16.7%)	1/29 (3.5%)	0.1498*
Number of patients with E. coli harbouring blaCTX−M−15		4/18 (22.2%)	24/29 (82.8%)	< 0.0001$
Number of patients with K. pneumoniae harbouring blaCTX−M−15		15/15 (100%)	26/26 (100%)	No calculation
Number of patients consistently colonised with one same E. coli ST		17/18 (94.4%)	16/29 (55.2%)	0.0042 $
Number of patients consistently colonised with one same K. pneumoniae ST		12/15 (80%)	17/26 (65.4%)	0.4799*
Number of patients with E. coli ST131		5/18 (27.8%)	16/29 (55.2%)	P = 0.0663$
Number of patients with K. pneumoniae ST25		11/15 (73.3%)	15/26 (57.7%)	P = 0.3166$

ESBLPE: extended-spectrum beta-lactamase-producing Enterobacterales, $ Chi-square test, * Fisher's exact test, ST: sequence type

There was no difference between patients in the short and long colonisation groups in the proportion who were colonised with E. coli or K. pneumoniae or in the proportion of patients colonised with E. coli and K. pneumoniae isolates of various STs, plasmid types and virulence factors.

Discussion

This prospective cohort study provides insights into the duration of intestinal colonisation for patients with a newly acquired ESBLPE and the potential impact of selected factors on this duration.

We found a median duration of ESBLPE colonisation for this cohort of 544 days. After one year and at the end of follow up, 61% and 23.4% of patients respectively remained ESBLPE colonised. Our findings are similar to those of previous hospital-based studies that have reported proportions of patients who remain ESBLPE colonised at 12 months of 22–44% [6–8]. By comparison, community-based studies assessing generally younger and healthier populations, such as those with travel-related ESBLPE acquisition, have found shorter durations of colonisation with 11% of patients remaining ESBLPE colonised at 12 months [35, 36]. These findings were confirmed by a recent systematic review and meta-analysis that found a faster decolonisation rate in travel related ESBLPE acquisition compared with discharged hospital patients residing in the community [37].

Using Cox regression analysis, we found that an increased intensity of hospital admissions and age ≥ 75 years were associated with prolonged ESBLPE colonisation. An increased intensity of antibiotic use was not associated with prolonged ESBLPE colonisation in this cohort. Given that an increased intensity of hospital admissions could reflect increased patient comorbidity, we included the Charlson comorbidity index score without inclusion of age in a revised Cox regression analysis. This showed that the Charlson comorbidity index score independent of age did not contribute to the duration of ESBLPE colonisation. A Spanish study found that multiple hospital readmissions was a risk factor for prolonged colonisation with ESBL-producing K. pneumoniae [38]. Other studies have not found that hospitalisation [8, 9] or age [6–9] was associated with prolonged ESBLPE colonisation. Studies have found that patient immobility [8] and immunosuppression [10] were associated with prolonged colonisation.

Consistent with previous studies, we did not find that patients’ comorbidity [6, 8], antibiotic use during follow up [6, 7, 9], place of residence [8] or travel to ESBL high-prevalence countries during follow up [6] contributed to the duration of ESBLPE colonisation.

Older age, intensity of hospitalisation, immobility and immunosuppression may all reflect senescence and in particular immune senescence which may contribute to prolonged ESBLPE colonisation. With aging, changes in the intestine [39], along with the aging immune system [40], may affect the gut microbiome. Disruption in the diversity and balance of the gut microbiota may affect its ability to prevent overgrowth of endogenous organisms as well as colonisation by exogenously introduced organisms [41]. It is possible that age-related gut dysbiosis may play a role in prolonged ESBLPE colonisation.

Several studies [42, 43] have found differences in the gut microbiome when comparing those with and without ESBLPE colonisation. These included low bacterial diversity, a lower abundance of health promoting genera and proliferation of potential pathogens in those who were ESBLPE colonised. Other studies [44, 45] conducted in generally younger and healthier populations did not find differences in the gut microbiome comparing those with or without ESBL-producing E. coli colonisation. Further research is required in this area.

We did not find any significant association between prolonged ESBLPE colonisation and any particular bacterial species, or colonisation with ESBL-producing E. coli or K. pneumoniae of certain STs, plasmid types or virulence factors. Other investigators [6, 7, 10] have also looked for similar associations without success. The relatively small number of sequenced isolates could have reduced our ability to identify any of these bacterial factors as factors associated with prolonged colonisation.

We found that E. coli harbouring bla_CTX−M−15 was associated with prolonged colonisation. It was the most common ESBL gene detected in this study, a consequence of its global spread [4]. Similarly, ESBL-producing E. coli harbouring CTX-M-group 9 were found to be associated with ESBLPE colonisation at 12 months in hospital patients [7] and with ESBLPE colonisation at 12 months in returned travellers [35]. Other investigators have not found an association between particular ESBL genes and prolonged colonisation [6, 10].

We found that colonisation with ESBL-producing E. coli phylogroup B2 was associated with prolonged colonisation. This has previously been found in adult patients [6, 7], infants [46] and returned travellers [47]. Twenty of 23 (86.9%) patients with prolonged colonisation who were colonised with this phylogroup were colonised with ExPEC lineages. ExPEC lineages were found to be more likely to persist compared with other E. coli lineages in a study of healthy travellers with very low antibiotic usage [48]. This finding supports the view that virulence factors required for invasive infection may also result in prolonged colonisation, as they may increase bacterial fitness within the normal gut environment [49]. We did not find that any individual E. coli ST or VF was associated with prolonged intestinal colonisation. This may suggest that the ESBL-producing E. coli phylogroup B2 strains in this cohort were more able to adapt to their surroundings with a combination of various bacterial characteristics resulting in increased bacterial fitness that contributed to a more prolonged duration of colonisation; this observation has also been made by other investigators [50].

Over two thirds of our patients with sequenced isolates, were colonised with the same strain of ESBL-producing E. coli (70.2%) or K. pneumoniae (70.7%) during the follow up period. This was accompanied with the carriage of (mostly) uniform resistance gene profiles and plasmid types. A study involving patients presenting to a tertiary care centre found persistent colonisation with the same ESBL-producing E. coli (83.6%) or K. pneumoniae species complex (100%) strain during the follow up period [10].

We found that an increased intensity of hospitalisation as well as colonisation with multiple ESBL-producing species and multiple E. coli STs was associated with prolonged ESBLPE colonisation. A possible explanation for these findings is that a more prolonged exposure to the hospital environment resulted in increased opportunities to acquire additional ESBL-producing species and strains, which subsequently contributed to a more prolonged ESBLPE colonisation. Other investigators have highlighted the importance of the hospital environment [51] as well as other sources of ongoing ESBLPE exposure such as residing in an acute care facility or international travel [52] that may increase the likelihood of co-colonisation with multiple ESBLPE species and/or replacement of one ESBLPE species with another, potentially leading to a more prolonged duration of colonisation. In addition to external (re)acquisition of ESBLPE, it is also known that plasmid transfer of ESBL genes can occur between different bacterial species and strains within the host [53]. We did not explore plasmid-mediated gene transfer to distinguish between endogenous long-term colonisation or an exogenous recolonization event. However, we observed that approximately half of the patients colonised with more than one K. pneumoniae or E. coli ST, or colonised with both species, had the same ESBL gene and at least one same plasmid type consistently detected in their isolates.

The strength of our study was the prospective nature of data collection and inclusion of patients from two hospitals. The main limitation was a relatively small number of included patients as well as the number of patients that did not complete some of the study requirements resulting in their exclusion from one or more components of the analysis. The relatively small number of patients will have reduced the power of the study to identify factors associated with prolonged colonisation.

Conclusion

We found that intestinal colonisation with a newly acquired ESBLPE commonly persists for one year or longer, most often with the same strain of E. coli or K. pneumoniae. We found older age, an increased intensity of hospitalisation, E. coli phylogroup B2, E. coli harbouring bla_CTX−M−15, colonisation with more than one species of ESBLPE and colonisation with more than one E. coli ST were associated with prolonged colonisation. It is very likely that combinations of multiple patient and bacterial factors influence the duration of ESBLPE colonisation. Factors such as older age, immune senescence, the composition of bacterial species in the gut and colonisation with (and re-exposure to) ESBLPE bacteria equipped with genetic determinants that increase their fitness and survival in the gut, may all contribute to prolonged colonisation. This prolonged duration of ESBLPE colonisation has implications for the possibility of community transmission once patients are discharged from hospital as well as hospital transmission at the time of potential future readmissions, and highlights the importance of infection control and public health measures to decrease the acquisition and spread of ESBLPE in hospitals and the community.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Study conception and design performed by: Dragana Drinković, David Holland, Lifeng Zhou, Simon Briggs, Susan Taylor, Hasan Bhally, Arlo Upton. Clinical data analysis and interpretation: Dragana Drinković, David Holland, Lifeng Zhou, Simon Briggs. Bioinformatics analysis and data interpretation: Scott Beatson. Statistical analysis: Lifeng Zhou. Initial manuscript draft: Dragana Drinković. All authors revised the final version of the manuscript.

Funding

This work was funded by Auckland Medical Research Foundation, Grant number 7112007.

Data availability

WGS data generated from this study are deposited in the NCBI Sequence Read Archive (Bioproject PRJNA1286807).

Code Availability

Not applicable.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval

The study received approval from the New Zealand Health and Disability Ethics Committee (NTY/10/03/027).

Consent to participate

Informed consent was obtained from all individual participants included in the study.

Consent for publish

Not applicable.