Comparison between IMP carbapenemase-producing Enterobacteriaceae and non-carbapenemase-producing Enterobacteriaceae: a multicentre prospective study of the clinical and molecular epidemiology of carbapenem-resistant Enterobacteriaceae

Abstract

Background

Carbapenem-resistant Enterobacteriaceae (CRE) are classified as carbapenemase-producing Enterobacteriaceae (CPE) and non-CPE; the majority of CPE in Japan produce IMP carbapenemase.

Objectives

We evaluated the clinico-epidemiological and microbiological information and effects of IMP-type carbapenemase production in CRE.

Methods

Patients with isolations of CRE (MICs of meropenem ≥2 mg/L, imipenem ≥2 mg/L or cefmetazole ≥64 mg/L) from August 2016 to March 2018 were included. Microbiological analyses and WGS were conducted and clinical parameters were compared between groups. Independent predictors for the isolation of CPE from patients were identified by logistic regression. For comparing clinical outcomes, a stabilized inverse probability weighting method was used to conduct propensity score-adjusted analysis.

Results

Ninety isolates (27 CPE and 63 non-CPE) were collected from 88 patients (25 CPE and 63 non-CPE). All CPE tested positive for IMP carbapenemase. Antibiotic resistance (and the presence of resistance genes) was more frequent in the CPE group than in the non-CPE group. Independent predictors for CPE isolation were residence in a nursing home or long-term care facility, longer prior length of hospital stay (LOS), use of a urinary catheter and/or nasogastric tube, dependent functional status and exposure to carbapenem. Although in-hospital and 30 day mortality rates were similar between the two groups, LOS after CRE isolation was longer in the CPE group.

Conclusions

IMP-CPE were associated with prolonged hospital stays and had different clinical and microbiological characteristics compared with non-CPE. Tailored approaches are necessary for the investigational and public health reporting, and clinical and infection prevention perspectives for IMP-CPE and non-CPE.

Introduction

Antimicrobial resistance poses a serious threat to public health and medical practice. Among various resistant pathogens, ESBL-producing Enterobacteriaceae (ESBLPE) and carbapenem-resistant Enterobacteriaceae (CRE) are of great importance given their ability to asymptomatically colonize the gut and their potential to spread through horizontal transmission via plasmids.

In contrast to ESBLPE, which were responsible for the CTX-M pandemic,¹ the molecular epidemiology of CRE depends on geographic location.^2,3 Most clinico-epidemiological information and its association with molecular characteristics has been reported for western countries, including those in North America and/or Europe, and have focused on molecular types such as KPC or NDM. However, IMP-type carbapenemases account for the majority of carbapenemases produced by Enterobacteriaceae in Japan and other Asian countries.^2,4,5

In Japan, infections with CRE have been classified as category V infectious diseases under the Infectious Diseases Control Law since 2014 and physicians who diagnose infections (not including colonization) caused by CRE must notify the government according to the National Epidemiological Surveillance of Infectious Diseases regulations.⁶ For reporting purposes, carbapenem resistance is defined as either resistance to meropenem (MIC ≥2 mg/L) or resistance to both imipenem (MIC ≥2 mg/L) and cefmetazole (≥64 mg/L), based on susceptibility tests at the reporting facility.⁶ The reported number of CRE infection cases in Japan based on this definition was approximately 1600 per year from 2015 to 2017 and approximately 2200 in 2018.^7,8 Microbiological analyses conducted on 865 CRE samples isolated in 2017 at the Public Health Institute revealed that only 239 (28%) isolates tested positive for a carbapenemase gene, among which the majority (227 isolates) were for IMP-type carbapenemase.⁵ These data suggest that the cases of CRE reported to the surveillance system were heterogeneous. For data acquisition from healthcare facilities for epidemiological information relevant to infection prevention and clinical management, it is therefore necessary to obtain more tailored information regarding CRE [i.e. carbapenemase-producing Enterobacteriaceae (CPE) and non-CPE] in Japan. However, IMP-type CPE (IMP-CPE) have not been isolated as frequently as other types of carbapenemase outside of Japan.^9,10 For example, a previous study in Japan found that the isolation frequency of CPE was low (0.35%: 17/4875 clinical isolates of Enterobacteriaceae).¹¹ Another study using clinical stool specimens from multiple institutes found 1.9% (5/268) CRE but no CPE.¹² Therefore, it is challenging to prospectively obtain the same level of detailed clinical and microbiological information on IMP-CPE as done for other molecular types.

In this prospective multicentre study, we aimed to evaluate the clinico-epidemiological and microbiological information and effects of carbapenemase production, particularly IMP-type carbapenemase production, among CRE in Japan.

Patients and methods

Ethics

All the study protocols were approved by the Human Research Ethics Committees at each facility. Written consent was obtained from all patients participating in this study.

Study settings and patients

We conducted a multicentre prospective cohort study at 11 tertiary care facilities in Japan. Patients with isolation of CRE from 1 October 2016 to 31 March 2018 were included in the study. CRE were defined as Enterobacteriaceae with either meropenem MIC ≥2 mg/L or imipenem MIC ≥2 mg/L and cefmetazole MIC ≥64 mg/L.⁶ For patients who had more than one episode of the same CRE isolation during the study period, only the first episode was analysed (i.e. the study included only unique-patient episodes). For patients with ≥2 CRE isolations on the same day, all isolates were included in the analyses. The characteristics of the healthcare facilities are summarized in Table S1 (available as Supplementary data at JAC Online). The number of inpatient beds at these facilities ranged from 425 to 927. Except for the Tokyo Metropolitan Cancer and Infectious Diseases Centre of Komagome Hospital (which specialized in patients with malignancy and infectious diseases, such as HIV infections) and the Shizuoka Cancer Centre and National Cancer Centre of Hospital East (which specialized in patients with malignancy), all institutes were tertiary care facilities with multiple departments to serve a wide variety of patients.

Clinical data collection

The following clinical data were collected: patient demographics; background conditions, comorbid conditions (including Charlson scores¹³ and age-adjusted Charlson scores¹⁴) and functional status; immunocompromised status; recent healthcare-associated exposure prior to admission, invasive procedures (such as endoscopy or surgery), stay in an ICU after admission before CRE isolation and presence of indwelling devices at the time of CRE isolation; exposure to antimicrobials in the past 1 month prior to culture; focus of infection (determined according to the National Healthcare Safety Network criteria¹⁵ by the Infectious Diseases consult service), clinical characteristics (CRE isolates were considered colonizers if patients did not have any sign of infection based on the above criteria and when they had asymptomatic bacteriuria), polymicrobial isolation and parameters regarding treatment; and outcomes, including in-hospital and 30 day mortality and length of hospital stay (LOS).

Microbiology

Standard identification and susceptibility testing of CRE was performed and interpreted in accordance with the CLSI criteria at microbiology laboratories at each facility.¹⁶ The carbapenem inactivation method (CIM) was used to screen carbapenemase production, as previously described.¹⁷ The transferability of the resistance genes associated with carbapenemase production was studied as described previously.¹⁸ CPE were defined as isolates that tested positive for both CIM tests and genes related to carbapenemase production, as identified by WGS, in this study.

WGS

WGS of all isolates was done using a HiSeq X system with Nextera XT library kits (Illumina, Tokyo, Japan) for the analysis of drug resistance genes and MLST, as previously described.¹⁹ Sequence typing of the plasmid replicons was performed as described elsewhere.^20,21 The obtained sequencing data were registered with the DNA Data Bank of Japan (DDBJ; accession no. DRA008386).

Statistical analysis

Univariate analyses were performed using Fisher’s exact tests or chi-squared tests for categorical variables and Mann–Whitney U-tests or Kruskal–Wallis tests for continuous variables. Multivariable models to determine predictors for the isolation of CPE as opposed to non-CPE were constructed using logistic regression.

For analysis of clinical outcomes for patients with CPE and non-CPE, propensity score adjustment was used to minimize baseline characteristic differences between both groups. We developed a non-parsimonious multivariable logistic regression model to estimate a propensity score for each patient’s likelihood of isolation of CPE as compared with non-CPE. The covariates included to generate the propensity score are listed in Table 5. Stabilized inverse probability weighting (IPW) was used to conduct the propensity score-adjusted analysis, as described previously.^22,23 The standardized bias was calculated, as described previously,²⁴ and the balance of each group was determined according to a standardized difference of less than 0.25 for all variables representing baseline characteristics. A logistic regression model (in-hospital and 30 day mortality) and linear regression model (on natural log-transformed LOS after isolation of CPE/non-CPE, excluding cases who died in hospital) using the stabilized IPW cohort were constructed. The effect estimates for LOS were reported as the multiplicative effects (antilog of the β coefficient).²⁵ The variable ‘infection (not colonization)’ was included in these models for additional adjustment.

Table 5.

Comparison of clinical parameters of CPE and non-CPE, univariate analysis^a

Variables	CPE (n = 25)	Non-CPE (n = 63)	P valueb
General patient demographics
Age, yearsc	77 (69–82)	70 (61–79)	0.043
Male sex	14 (56)	39 (61.9)	0.636
Travel outside of Japan within 1 year	4 (16)	3 (5.3)	0.192
Healthcare-associated exposure prior to admission
Nursing home or LTCF residence	6 (24)	4 (6.3)	0.028
Hospitalization in the past 3 months	14 (56)	25 (39.7)	0.234
IV therapy or haemodialysis or chemotherapy in the past 1 month	13 (52)	32 (51.6)	>0.999
GI tract endoscopy or surgery in the past 1 month	8 (32)	15 (23.8)	0.433
Any healthcare-associated exposure prior to admission	18 (72)	40 (63.5)	0.619
Healthcare-associated exposure prior to CPE/non-CPE isolation after admission
ICU stay	10 (40)	25 (41.7)	>0.999
GI tract endoscopy	6 (27.3)	6 (10.2)	0.078
Surgery	8 (36.4)	23 (39)	>0.999
LOS before isolation of CPE/non-CPE, daysc	24 (13-72)	14 (3-25)	0.01
Presence of devices at the time of CPE/non-CPE isolation
Central line	11 (44)	15 (23.8)	0.074
Peripheral line	14 (56)	39 (61.9)	0.636
Urinary catheter	17 (68)	23 (36.5)	0.009
Tracheostomy tube	4 (16)	8 (12.7)	0.735
Intubation tube	1 (4)	4 (6.3)	>0.999
Gastrostomy tube	1 (4)	3 (4.8)	>0.999
Nasogastric tube	11 (44)	7 (11.1)	0.001
Drain tubed	6 (24)	7 (11.1)	0.181
Any device	24 (96)	54 (85.7)	0.270
Acute and chronic conditions on admission
Dependent functional status	19 (76)	24 (38.1)	0.002
Congestive heart failure	3 (12)	8 (12.7)	>0.999
Cerebrovascular accident	7 (28)	11 (17.5)	0.379
Dementia	4 (16)	4 (6.3)	0.216
Chronic pulmonary disease	0 (0)	7 (11.1)	0.184
Connective tissue disease	2 (8)	5 (7.9)	>0.999
Diabetes mellitus (uncomplicated)	7 (28)	5 (7.9)	0.033
Liver disease (mild)	3 (12)	0 (0)	0.021
Renal diseases (moderate to severe)	0 (0)	7 (11.1)	0.184
Leukaemia or lymphoma	2 (8)	3 (4.8)	0.620
Localized solid tumour	8 (32)	15 (23.8)	0.433
Metastatic solid tumour	4 (16)	4 (6.3)	0.216
Charlson combined condition scorec	3 (2–5)	2 (1–3)	0.120
Age-adjusted Charlson combined condition scorec	6 (4–8)	5 (4–6)	0.030
Any immunosuppressive statuse	2 (8)	7 (11.1)	>0.999
Antimicrobial exposure in the past 1 month
Penicillin (excluding BLBLI)	2 (8)	1 (1.6)	0.193
Penicillin (only BLBLI)	13 (52)	21 (33.3)	0.145
Cephalosporin (excluding cephamycins)	10 (40)	24 (38.1)	>0.999
Cephamycin	2 (8)	8 (12.7)	0.718
Carbapenem	10 (40)	9 (14.3)	0.019
Fluoroquinolone	3 (12)	2 (3.2)	0.136
Glycopeptide	2 (8)	6 (9.5)	>0.999
Any antibiotics	22 (88)	50 (79.4)	0.541
Clinical characteristics
Infection (not colonization)	11 (44)	22 (34.9)	0.470
Polymicrobial isolationf	19 (76)	48 (76.2)	>0.999
Bacteraemia due to CPE/non-CPE	4 (16)	3 (4.8)	0.097
Focus of infectiong
Urinary tract infection	2 (8)	7 (11.1)	>0.999
Pneumonia	3 (12)	5 (7.9)	0.683
Intra-abdominal	4 (16)	4 (6.3)	0.216
Skin and soft tissue infection (including surgical site infection)	1(4)	5 (7.9)	0.670
Severity of infectiong
Pitt bacteraemia score,10 median (range)	2 (0–3)	2 (0–4)	0.853
quick-SOFA11c	2 (1–2)	1 (1–2)	0.639
ICU stay at the time of isolation	2 (18.6)	3 (13.6)	>0.999
Treatmentg
Effective empirical therapyh	3 (27.3)	15 (68.2)	0.061
Inadequate source controli	0	2 (9)	0.542
Outcome
LOS after isolation of CPE/non-CPE excluding cases who died in hospital, daysc	46 (18–72)	29 (13–45)	0.091
LOS after isolation of CPE/non-CPE, daysc	40 (18–72)	29 (13–47)	0.143
Total LOS, daysc	83 (41–150)	45 (30–73)	0.017
In-hospital mortality	4 (16)	5 (8.1)	0.272
30 day mortality	3 (12.5)	3 (5.1)	0.349

Abbreviations: BLBLI, β-lactam/β-lactamase inhibitor combinations; GI, gastrointestinal; LTCF, long-term care facility.

a

Data are presented as n (%) unless indicated otherwise.

b

Bold indicates statistically significant results (P < 0.05).

c

Values shown are median (IQR).

d

Drain tubes included abdominal drain tubes and chest drain tubes.

e

Including one or more of the following: neutropenia (<500 neutrophils/μL) at the time of culture, glucocorticoid/steroid use (doses greater than or equal to an equivalent of 20 mg of prednisone per day for at least 1 month), chemotherapy or immunosuppressant use (such as anti-TNF-α therapy, anti-IL-6 receptor/anti-CD20 monoclonal antibodies, selective T cell costimulation blocker, methotrexate) in the past 1 month, post-transplantation in the past 3 months or HIV infection.

f

Isolation of additional bacteria other than CPE/non-CPE from the same culture.

g

Included only the patients with infection (not colonization). Pitt bacteraemia score was calculated only for bacteraemic cases. Numbers (and percentages) were calculated accordingly.

h

Effective empirical therapy was defined as antibiotics given prior to obtaining susceptibility results regarding which CPE/non-CPE were susceptible based on CLSI criteria¹⁶ including carbapenems.

i

Inadequate source control included undrained abscess.

All P values were two-sided and P values of less than 0.05 were considered statistically significant. Throughout the text, the percentages displayed are the ‘valid percentages’, which indicate the percentage excluding the missing data from the denominator. Statistical analyses were performed with IBM SPSS Statistics 25 (2017).

Results

Summary of the study cohort

During the study period, 90 isolates (27 CPE and 63 non-CPE) were collected from 88 patients (25 with CPE and 63 with non-CPE) in 9 of 11 hospitals (Table S1). Two hospitals that participated in this study did not report CRE isolation during the study period. Two isolates (one Enterobacter cloacae and one Serratia marcescens) were positive for CIM but were not found to carry any carbapenemase genes and were therefore categorized as non-CPE.

Bacterial species and isolation sites

Bacterial species and isolation sites of CRE are summarized in Table 1. CPE included 10 (37%) E. cloacae, 6 (22.2%) Klebsiella pneumoniae, 4 (14.8%) Escherichia coli, 3 (11.1%) Citrobacter freundii, 2 (7.4%) Klebsiella oxytoca and 1 (3.7%) each of Enterobacter aerogenes and S. marcescens. Non-CPE included 34 (54%) E. aerogenes, 15 (23.8%) E. cloacae, 4 (6.3%) each of E. coli, K. pneumoniae and S. marcescens, and 2 (3.2%) C. freundii. Both CPE and non-CPE were most commonly isolated from sputum [CPE/non-CPE: 11 (40.7%)/25 (39.7%)], followed by urine [7 (25.9%)/12 (19%)], intra-abdominal samples [3 (11.1%)/7 (11.1%)] and bile [3 (11.1%)/6 (9.5%)]. Four (14.8%) CPE and three (4.8%) non-CPE were isolated from blood.

Table 1.

Bacterial species and isolation sites of CRE from participating facilities

	Total, n (%)	E. cloacae	E. aerogenes	E. coli	C. freundii	K. pneumoniae	K. oxytoca	S. marcescens
CPEa (n = 27)
total, n (%)		10e (37)	1 (3.7)	4 (14.8)	3f (11.1)	6 (22.2)	2 (7.4)	1g (3.7)
blood	4 (14.8)	2			1	1
sputumb	11 (40.7)	4			1	4	1	1
urine	7 (25.9)	2	1	1	2	1
bile	3 (11.1)			2			1
intra-abdominalc	3 (11.1)	2			1
woundd	1 (3.7)			1
stool	3 (11.1)	1			1			1
Non-CPE (n = 63)
total, n (%)		15 (23.8)	34h (54)	4 (6.3)	2 (3.2)	4 (6.3)	0	4 (6.3)
blood	3 (4.8)	1	1			1
sputum	25 (39.7)	6	17					2
urine	12 (19)	1	9	1				1
bile	6 (9.5)	4	2
intra-abdominal	7 (11.1)	1	2	2	1			1
wound	6 (9.5)	2	2		1	1
stool	3 (4.8)			1		2

a

Including four isolates from two patients from whom two CPE were isolated from same-day samples. Case 1: isolation from intra-abdominal drain (E. cloacae and C. freundii); Case 2: isolation from blood (K. pneumoniae) and bile (E. coli).

b

Including one bronchoalveolar lavage fluid and one lung specimen.

c

Including specimens such as intra-abdominal abscess and ascites.

d

Including specimens from skin and soft tissue infection or decubitus ulcers.

e

One isolate was simultaneously isolated from blood and intra-abdominal samples.

f

One isolate was simultaneously isolated from blood, urine, sputum and stool.

g

One isolate was simultaneously isolated from sputum and stool.

h

One E. aerogenes was isolated from the anterior nares.

Antimicrobial susceptibility profiles

The susceptibility profiles of CRE included in this study are summarized in Table 2. Among non-β-lactam antibiotics, levofloxacin had lower susceptibility rates in CPE [22 (81.5%)] than in non-CPE [58 (92.1%)], although this difference was not significant. Among β-lactams, meropenem [CPE/non-CPE: 0/48 (76.2%)], ampicillin/sulbactam [0/31 (49.2%)], piperacillin/tazobactam [11 (40.7%)/51 (81%)], ceftazidime [0/41 (65.1%)] and cefepime [9 (33.3%)/51 (81%)] had significantly lower susceptibility rates in CPE than in non-CPE.

Table 2.

Susceptibility profiles of CRE included in this study

	Isolates susceptible,an (%)
Antibiotic	total	CPE (n = 27)	non-CPE (n = 63)	P valueb (CPE versus non-CPE)
Levofloxacin	80 (88.9)	22 (81.5)	58 (92.1)	0.159
Gentamicinc	63 (86.3)	19 (82.6)	44 (88)	0.72
Amikacin	87 (96.7)	26 (96.3)	61 (96.8)	>0.999
Imipenemd,e	6 (8.2)	1 (8.3)	5 (8.2)	>0.999
Meropeneme	48 (53.3)	0 (0)	48 (76.2)	<0.001
Ampicillin/sulbactam	31 (34.4)	0 (0)	31 (49.2)	<0.001
Piperacillin/tazobactam	62 (68.9)	11 (40.7)	51 (81)	<0.001
Cefmetazolef	1 (1.1)	0 (0)	1 (1.6)	>0.999
Ceftazidime	41 (45.6)	0 (0)	41 (65.1)	<0.001
Aztreonam	62 (68.9)	19 (70.4)	43 (68.3)	>0.999
Cefepime	60 (66.7)	9 (33.3)	51 (81)	<0.001
Trimethoprim/sulfamethoxazoled	61 (83.6)	11 (91.7)	50 (82)	0.676

a

Including susceptible and intermediate isolates, based on CLSI criteria¹⁶ unless otherwise noted.

b

Bold text indicates statistically significant results (P < 0.05).

c

Data not available for isolates from Japanese Red Cross Narita Hospital (n = 16; 3 CPE and 13 non-CPE).

d

Data not available for isolates from National Centre for Global Health and Medicine (n = 17; 15 CPE and 2 non-CPE).

e

Isolates with MIC = 2 mg/L were categorized as resistant.

f

Seventeen isolates from the National Center for Global Health and Medicine with MIC ≥ 32 mg/L were categorized as resistant.

Resistance genes and transferability by conjugation

Table 3 shows data of resistance genes of CRE included in this study. All CPE were positive for IMP-type carbapenemase, including 11 (40.7%) IMP-11, 6 (22.2%) IMP-42, 4 (14.8%) IMP-6 and 3 (11.1%) each of IMP-10 and IMP-1. Overall, resistance genes were found more commonly in the CPE group than in the non-CPE group. Except for bla_TEM1B, which was found more frequently in the CPE group than in the non-CPE group (P < 0.001), the prevalence of other β-lactamase genes varied and they were found in both CPE and non-CPE groups. The prevalence of resistance genes related to aminoglycoside resistance was different between the CPE and non-CPE groups; aac(3)-IId, aac(6′)-IIc and aac(6′)-Ia were more common in the CPE group than in the non-CPE group (P = 0.049, P < 0.001 and P < 0.001, respectively). Similarly, fluoroquinolone-related resistance genes were more frequent in the CPE group than in the non-CPE group (qnrS1: P = 0.024, qnrB6: P < 0.001). Similar trends were observed for sulphonamide-, trimethoprim-, tetracycline- and fosfomycin-related resistance genes as well as efflux pump-related genes, such as oqxA and oqxB.

Table 3.

Resistance genes of CRE included in this study

Values shown are n (%) of CPE/non-CPE. A dash indicates our analysis could not determine the specific type of fosA.

^aComparison of CPE group versus non-CPE group showed statistically significant differences in prevalence (P < 0.05).

The transferability of resistance genes by conjugation was examined in all 90 CPE isolates positive for bla_IMP. Eighteen (67%) isolates harbouring bla_IMP exhibited transferability of resistance genes by conjugation (six K. pneumoniae, five E. cloacae, three E. coli, two C. freundii and one each of E. aerogenes and K. oxytoca), with a frequency of 5.9 × 10⁻⁶ to 1.7 × 10⁻³. The transferability of each bla_IMP subtype is summarized in Table S2. The transferability did not differ significantly among different IMP subtypes in terms of proportion of transferable isolates or mean frequency.

Distribution of MLST and bla_IMP subtypes in each facility

To investigate the clonality of the isolates, we analysed the distribution of the MLST and bla_IMP subtypes in each facility (Table 4). Three carbapenemase-producing (CP) K. pneumoniae isolates from a single facility (Facility 1) belonged to the same MLST (ST268) and possessed the same IMP-subtype (IMP-11). Eleven isolates of different bacterial species possessed bla_IMP-11 in Facility 1 and thus sequence typing of plasmid replicons was also performed on these isolates (Table 4). All eleven isolates possessed plasmid replicons of the IncL/M incompatibility group, which might encode bla_IMP-11. Other than these in Facility 1, there were no findings to suggest the clonal spread of CPE or non-CPE.

Table 4.

bla_IMP subtype, MLST and plasmid replicon type of isolates

blaIMP	Facility	Species	STa	Replicon type (incompatibility group)
CPE
1	1	K. pneumoniae	268	H	FIA	FIB	FII
1	4	K. pneumoniae	12		FIA	FIB	FII	L/M
1	7	S. marcescens	ND	H
6	1	E. cloacae	78	H
6	2	E. cloacae	133	H					A/C
6	6	E. coli	131		FIA	FIB	FII		N
6	7	E. cloacae	133	H
10	2	E. cloacae	742	H
10	4	C. freundii	125					L/M
10	4	K. oxytoca	ND			FIB		L/M
11	1	E. cloacae	148			FIB	FII	L/M
11	1	E. cloacae	88	H		FIB	FII	L/M	R	P
11	1	E. cloacae	74		FIA	FIB		L/M	R
11	1	E. coli	1193					L/M
11	1	E. coli	7508					L/M
11	1	E. coli	131		FIA	FIB		L/M	R
11	1	K. oxytoca	ND			FIB	FII	L/M
11	1	K. pneumoniae	1303			FIB		L/M
11	1	K. pneumoniae	268		FIA	FIB		L/M	R
11	1	K. pneumoniae	268		FIA	FIB		L/M	R
11	1	K. pneumoniae	268		FIA	FIB		L/M	R
42	1	C. freundii	ND		FII	FIB
42	1	E. cloacae	484			FIB	FII
42	2	E. cloacae	133	H
42	4	E. aerogenes	ND
42	4	E. cloacae	125		FIA	FIB	FII	L/M
42	7	C. freundii	62		FIA			L/M
Non-CPE
	1	E. aerogenes	ND
	1	S. marcescens	ND
	2	C. freundii	ND
	2	E. aerogenes	9
	2	E. aerogenes	135
	2	E. aerogenes	ND (9)
	2	E. cloacae	162
	3	E. aerogenes	12
	3	E. aerogenes	115
	3	E. aerogenes	190
	3	E. aerogenes	ND (7)
	3	E. cloacae	116
	3	E. cloacae	ND
	3	S. marcescens	ND
	4	E. aerogenes	ND (2)
	4	E. cloacae	466
	4	E. cloacae	ND (2)
	4	E. coli	131
	4	E. coli	720
	4	K. pneumoniae	ND
	5	C. freundii	ND
	5	E. aerogenes	ND (3)
	5	E. cloacae	56
	5	E. cloacae	ND (5)
	5	S. marcescens	ND
	6	E. aerogenes	103
	6	E. aerogenes	151
	6	E. aerogenes	ND
	6	E. cloacae	ND (2)
	6	E. coli	131
	6	E. coli	405
	6	K. pneumoniae	25
	6	K. pneumoniae	147
	6	S. marcescens	ND
	7	K. pneumoniae	290
	8	E. aerogenes	173
	8	E. aerogenes	ND
	8	E. cloacae	ND
	9	E. aerogenes	135
	9	E. aerogenes	ND

ND, not determined.

a

Number of isolates of each ST is one, unless shown in parentheses.

Prevalence of ESBL/plasmid-mediated AmpC (pAmpC) and porin loss/mutations among non-CP E. coli and K. pneumoniae

Carbapenem resistance among non-CPE is reportedly due to a combination of porin loss and the production of AmpC or ESBL.^26–28 We conducted analyses of the prevalence of ESBL/pAmpC and porin loss/mutation in E. coli, which usually has low-level expression of chromosomal AmpC β-lactamases, and K. pneumoniae, which does not possess chromosomal AmpC β-lactamases (Table S3).²⁹ Three out of four E. coli isolates possessed the ESBL gene, all four isolates lacked OmpC genes and one isolate had a frameshift mutation of the OmpF gene. All four K. pneumoniae isolates were positive for the ESBL gene and one isolate also had a pAmpC gene. One K. pneumoniae isolate possessed a nonsense mutation of the OmpK35 gene and a missense mutation of the OmpK36 gene; other K. pneumoniae isolates had missense mutations of at least one of the two (OmpK35 or OmpK36) genes.

Comparison of baseline clinical parameters of CPE and non-CPE

Further analyses of clinical epidemiology were conducted on 88 patients (25 CPE and 63 non-CPE). The results of univariate analyses comparing patients with CPE with patients with non-CPE are shown in Table 5. Patients with CPE were more likely to be older, dependent in terms of functional status and to have resided in a nursing home or long-term care facility prior to admission. Device use, such as urinary catheters and nasogastric tubes, was more frequent in the CPE group than in the non-CPE group. Moreover, patients with CPE tended to have undergone gastrointestinal endoscopy more often than those with non-CPE. The median Charlson scores were slightly higher in patients with CPE than in patients with non-CPE; however, the proportion of patients with immunosuppressive status was similar between the two groups. Prior carbapenem exposure was significantly higher in the CPE group than in the non-CPE group.

Next, we conducted multivariate analyses to identify independent predictors for the isolation of CPE compared with non-CPE (Table 6). Because the results of univariate analyses suggested that cases with CPE were more elderly and had more frequent comorbidities, and these characteristics could be confounders of other potential predictors (such as antimicrobial exposure, device use and healthcare exposure), each potential predictor variable was analysed by controlling for the confounding effects of an age-adjusted Charlson score. Independent predictors for the isolation of CPE compared with non-CPE were residence in a nursing home or long-term care facility, LOS before isolation of CPE/non-CPE ≥28 days (analysis using LOS as a continuous variable was also significant; P = 0.014), use of a urinary catheter and/or nasogastric tube at the time of CPE/non-CPE isolation, dependent functional status and exposure to a carbapenem in the past month.

Table 6.

Multivariate analyses of predictors for isolation of CPE

Variables	CPE cases versus non-CPE cases
	aOR (95% CI)a	P valuea
Nursing home or LTCF residence	4.18 (1.04–16.81)	0.044
LOS before isolation of CPE/non-CPE ≥28 days	3.82 (1.34–10.87)	0.012
Urinary catheter at the time of CPE/non-CPE isolation	3.37 (1.23–9.25)	0.019
Nasogastric tube at the time of CPE/non-CPE isolation	6.73 (2.09–21.7)	0.001
Dependent functional status	5.16 (1.74–15.29)	0.003
Exposure to carbapenem in the past 1 month	4.04 (1.32–12.44)	0.015

Abbreviations: aOR, adjusted OR; LTCF, long-term care facility.

a

Controlled for the confounding effects of age-adjusted Charlson combined condition score for each variable of potential predictor listed in Table 5 (only statistically significant results were presented).

Comparison of clinical characteristics and outcomes of CPE and non-CPE

There were no significant differences in clinical characteristics between the two groups, including the proportion of colonization (not infection) and polymicrobial isolation (Table 5). Bacteraemia was slightly more frequent in the CPE group; however, the severity of infection was not different between the two groups. The lack of significance of these parameters did not change when comparing the two groups in the stabilized IPW cohort. Patients in the non-CPE group received effective empirical therapy more frequently than patients in the CPE group, although this difference was not significant in univariate analysis [CPE versus non-CPE: n = 3 (27.3%) versus 15 (68.2%), P = 0.061]. The difference became smaller in the stabilized IPW cohort [n = 4 (50%) versus 14 (66.7%), P = 0.433]. Inadequate focus removal was infrequent and statistically similar in both groups. A carbapenem (meropenem) was administered as the only effective empiric therapy to two patients with non-CPE [meropenem MIC ≤1 mg/L (n = 2)], neither of whom died in the hospital. Fluoroquinolones, such as levofloxacin or ciprofloxacin, were most commonly used as definitive therapy, particularly in the CPE group (n = 13; n = 8 in CPE and n = 5 in non-CPE). A carbapenem was used as the only effective definitive therapy in six cases in the non-CPE group but no cases in the CPE group. None of these six cases died in hospital.

In the univariate analyses, the in-hospital and 30 day mortality rates were statistically similar between the two groups, although they tended to be higher in the CPE group (Table 5). Among cases with infection (not colonization), trends toward higher mortality in the CPE group than in the non-CPE group were noted, although these differences were not statistically significant [in-hospital mortality: CPE (n = 4, 36.4%) versus non-CPE (n = 2, 9.1%), P = 0.146; 30 day mortality: CPE (n = 3, 27.3%) versus non-CPE (n = 1, 5.3%), P = 0.126]. Total LOS was longer in the CPE group, whereas LOS after isolation of CPE/non-CPE was not significantly different between the two groups, even after excluding patients who died in the hospital (although ‘LOS after isolation’ tended to be longer in the CPE group). The results of analyses limited to the ‘infection (not colonization)’ cohort were similar (P = 0.693 for LOS after isolation, P = 0.198 for LOS after isolation excluding cases of death). None of the outcome parameters differed among facilities (P = 0.429/0.249 for in-hospital/30 day mortality; P = 0.207/0.635/0.801 for total LOS/LOS after isolation/LOS after isolation excluding cases of death).

In-hospital death and 30 day mortality rates were similar between the CPE and non-CPE groups in the propensity score-adjusted cohort (Table 7). However, LOS after isolation of CPE/non-CPE was 1.93 times longer in the CPE group than in the non-CPE group, after adjusting for infection (not colonization; 95% CI 1.13–3.31; P = 0.018).

Table 7.

Propensity score-adjusted analyses^a of clinical outcomes of CPE as compared with non-CPE

Variable	CPE	Non-CPE	Adjusted ORb (95% CI)	P value
In-hospital mortality	2 (10.5%)	8 (11.8%)	1.07 (0.22–5.21)	0.934
30 day mortality	1 (5.6%)	2 (3.1%)	1.59 (0.17–15.09)	0.688
LOS after isolation of CPE/non-CPE, daysc	58 (32–78)	29 (16–42)	1.93 (1.13–3.31)	0.018

a

Covariates included to generate the propensity score were the following: age-adjusted Charlson combined condition score, leukaemia or lymphoma, dependent functional status, use of nasogastric tube, exposure to a carbapenem within 1 month, polymicrobial isolation and history of admission in the past 3 months. Propensity score adjustments were performed using the stabilized IPW as described in the Patients and methods section. Maximum/median (IQR) standardized bias: (i) unadjusted [0.83/0.21 (0.11–0.37)]; (ii) propensity score-adjusted [0.19/0.08 (0.03–0.11)].

b

Controlled for variable of ‘infection (not colonization)’. OR for in-hospital death and 30 day mortality. The effect estimate for LOS after isolation of CPE/non-CPE, excluding cases who died in hospital, was reported as the multiplicative effect (the antilog of the β coefficient).

c

Excluding cases who died in hospital.

Although active surveillance, such as perirectal/stool swabs, was not conducted in this study, repeated isolation of the same CPE/non-CPE from clinical samples was observed in nine patients, of which three were from different sites than previous isolations. The duration from the first isolation to the subsequent isolation ranged from 10 to 207 days (three CPE: 30–207 days, six non-CPE: 10–80 days). Only the first isolates were investigated for resistance genes; thus, microbiological analyses on the subsequent isolates were conducted only at each facility.

Discussion

In this study, we report novel findings on the characteristics of CRE from multiple tertiary care hospitals. These include the molecular microbiological characteristics of IMP-CPE and non-CPE, as well as clinical epidemiology and its differences between IMP-CPE and non-CPE, and predictors for isolation and outcomes.

Risk factors for the isolation of IMP-CPE have been reported previously, including antimicrobial exposure, longer stays in hospital and device use, although these data were from a single institution, evaluated only one species and used a cross-sectional design.^30–32 Findings from our study provide valuable information from multiple centres in Japan, such as exposure history to carbapenems, admission in long-term care facilities/nursing homes, long hospital stays and device use (e.g. urinary catheters or nasogastric tubes) to predict the isolation of CPE compared with non-CPE.

In our study, mortality was similar between the CPE and non-CPE groups after carefully controlling for differences in baseline characteristics using propensity score-adjusted analyses. A previous study from North America, which compared mortality rates in patients with bacteraemia due to CP-CRE and non-CP-CRE, found higher mortality in the CP-CRE group.³³ Their study only included cases of bacteraemia and the majority (>90%) of carbapenemases were KPC-type. Moreover, resistance rates against multiple classes of antibiotics were much higher than those observed in our IMP-CPE cohort. These differences could explain the inconsistencies between their results and ours. In our study, the isolation of CPE was associated with an LOS almost twice as long as that associated with the isolation of non-CPE, even after excluding patients who died in the hospital. The LOS among participating facilities after isolation of CRE did not significantly differ; thus, it is unlikely that the results were driven by differences in protocol or practice at each facility. The reason for the longer stay in the CPE group was unclear; however, this was an important finding suggesting that the isolation of CPE affected patient outcomes to a greater extent than non-CPE in our multicentre cohort of CRE.

Although identifying the detailed mechanisms of non-CP-CRE is beyond the scope of this study, the analyses on E. coli and K. pneumoniae suggested that the production of ESBL combined with the loss/mutation of porin contributed to carbapenem resistance among non-CP-CRE. This finding is in accordance with previous studies³⁴ and similar mechanisms involving AmpC β-lactamase and porin loss likely contribute to carbapenem resistance among other non-CP-CRE with the ability of chromosomal AmpC production.^26–28

Two CIM-positive isolates (one E. cloacae and one S. marcescens) were negative for any carbapenemase genes using ResFinder for WGS data.³⁵ These isolates also tested positive in modified CIM tests.³⁶ In addition to the analyses using WGS data, we conducted PCR to detect carbapenemase genes (bla_SPM, bla_AIM, bla_DIM, bla_GIM, bla_SIM, bla_BIC, bla_SME, bla_IMI, bla_NMC-A and bla_GES), as previously described,^37,38 and none of the carbapenemase genes tested positive (data not shown). Based on these data, we considered these two isolates to be falsely positive in the CIM test as no known carbapenemases were detected. False-positive results in a CIM test with E. cloacae were reported previously.³⁹ These two isolates were both positive for AmpC β-lactamase (E. cloacae: bla_MIR-6; S. marcescens: bla_ACT-12 and bla_CMY-8) and this might have affected the CIM test results.

As a limitation, we might have missed some CPE such as OXA-type CPE by using the screening criteria we used in our study. However, our screening criteria were in accordance with the reporting criteria for the Infectious Diseases Control Law. Currently, it is not feasible for all microbiology laboratories inside hospitals to determine low MICs (<1 mg/L) for carbapenems, or to conduct rapid molecular analyses. For non-carbapenem-resistant CPE, further studies using stricter MIC criteria are necessary.

Our study results reveal that IMP-CPE and non-CPE were not only microbiologically but also clinically distinct; thus, the combined data (as ‘CRE data’) would be of minimal use to further explore appropriate methods for prevention of infection and for treatment. For example, IMP-CPE-associated outbreaks in healthcare facilities have been reported; ^30,31,40 however, to the best of our knowledge, no similar outbreaks due to the horizontal transmission of non-CP-CRE have been reported in Japan. The CDC recommends contact precautions for both CP-CRE and non-CP-CRE, with a potential requirement for more aggressive infection control interventions for CP-CRE.⁶ Studies using data that distinguish between CP-CRE and non-CP-CRE may provide further insight into these issues. Because resistance to carbapenems, particularly to imipenem, even when combined with resistance to cefmetazole, did not readily predict the production of IMP-carbapenemase, rapid identification methods that are feasible to perform at each facility level would be useful for initiating prompt management including appropriate empirical therapy and infection control of IMP-CPE.

In conclusion, IMP-CPE were associated with prolonged hospital stays and had different clinical and microbiological characteristics than non-CPE. Tailored approaches are necessary for the investigational and public health reporting, and clinical and infection prevention perspectives for IMP-CPE and non-CPE.

Funding

This work was supported by a Grant-in-Aid for Young Scientists (B) (grant no. 16K21652)(K.H.); a Grant-in-Aid for Scientific Research (C) (grant no. 17K10027)(R.N.); a grant for International Health Research from the Ministry of Health, Labor and Welfare of Japan (grant no. 19A1022)(K.H.); and Japan Agency for Medical Research and Development (AMED) (grant no. JP19fm0108001) from the Japan Initiative for Global Research Network on Infectious Diseases (T.M-A.).

Transparency declarations

None to declare.