Household transmission of carbapenemase-producing Enterobacteriaceae: a prospective cohort study

Abstract

Objectives

To estimate the transmission rate of carbapenemase-producing Enterobacteriaceae (CPE) in households with recently hospitalized CPE carriers.

Methods

We conducted a prospective case-ascertained cohort study. We identified the presence of CPE in stool samples from index subjects, household contacts and companion animals and environmental samples at regular intervals. Linked transmissions were identified by WGS. A Markov model was constructed to estimate the household transmission potential of CPE.

Results

Ten recently hospitalized index patients and 14 household contacts were included. There were seven households with one contact, two households with two contacts, and one household with three contacts. Index patients were colonized with bla_OXA-48-like (n = 4), bla_KPC-2 (n = 3), bla_IMP (n = 2), and bla_NDM-1 (n = 1), distributed among divergent species of Enterobacteriaceae. After a cumulative follow-up time of 9.0 years, three family members (21.4%, 3/14) acquired four different types of CPE in the community (hazard rate of 0.22/year). The probability of CPE transmission from an index patient to a household contact was 10% (95% CI 4%–26%).

Conclusions

We observed limited transmission of CPE from an index patient to household contacts. Larger studies are needed to understand the factors associated with household transmission of CPE and identify preventive strategies.

Introduction

The WHO has classified carbapenem-resistant Enterobacteriaceae (CRE) as a priority pathogen in critical need of new antibiotics.¹ In the absence of safe and effective antimicrobials, prevention of acquisition remains the primary intervention, since colonization with CRE is associated with a greatly increased risk of invasive infection with these organisms.² Existing infection prevention and control strategies that focus on interrupting transmission pathways, with an emphasis on early identification and physical separation of CRE carriers, have produced some success.³ However, those who do acquire CPE continue to be colonized for up to 6 months after hospital discharge, with 30% carrying it beyond 1 year,⁴ providing an avenue for community spread of the pathogen. The aim of this study was to estimate the transmission rate of CPE in households with recently hospitalized CPE carriers.

Methods

We conducted a case-ascertained cohort study⁵ from October 2016 to August 2017 in Singapore. Newly identified CPE-colonized patients (index subjects) were recruited from two hospitals. Inclusion criteria for index subjects were: age ≥21 years, current residence in a household with at least one other person, and able to give consent. Consenting household members (contacts) ≥21 years old were recruited. The primary outcome was a clinical transmission event defined as the detection of a carbapenemase gene identical to that of the index patient in family members either by culture method or stool microbiome analysis.

We collected stool samples from index subjects, household contacts and companion animals (if any) and environmental samples at baseline (time of recruitment) (S0), weekly for 4 weeks (S1–S4), monthly for 5 months (S5–S9) and bimonthly for 6 months (S10–S12) (Figure S1, available as Supplementary data at JAC Online). Baseline samples for household contacts, companion animals, and environmental samples were collected before the index subjects were discharged home. Index subjects and household contacts answered a short questionnaire via face-to-face or telephone interview at every visit.

All stool samples were collected by the study subjects in a sterile stool container. Household environmental samples were collected by the study team at every visit using Cultiplast^® Tampone Swabs with Amies transport media. Sites sampled were door handle, living room couch, bed, toilet flush handle, kitchen sink, and the chopping board. All samples were processed within 48 h for identification of CRE and carbapenemase-producing Enterobacteriaceae (CPE).

DNA was extracted from stool samples for metagenomic studies with a manual QIAamp DNA Stool Mini Kit (Qiagen), as per the manufacturer’s instructions, and stored at −80°C. DNA was extracted from pure subcultures of CPE using the MagNA Pure Compact platform (Roche Applied Science, Germany) as per manufacturer’s instructions for WGS. All CPE isolates and stool DNA were sequenced using an Illumina HiSeq 4000 sequencer. Full clinical, microbiological, and genomic methods are available in the Supplementary data.

We compared the baseline characteristics of index patients and household contacts with Fisher’s exact test for categorical variables and Wilcoxon rank-sum test for continuous variables. We analysed the data with a 4-state continuous time Markov model to assess transmission dynamics in terms of carriage acquisition and loss while accounting for interval censoring (Figure S2). The states were: (i) non-colonized; (ii) index patient colonized and household member non-colonized; (iii) index patient non-colonized and household member colonized; and (iv) both colonized (see Supplementary data). Analysis was performed using the msm package in R 3.4.4.⁶

Ethics

This study was approved by the National Health Group’s institutional review board (DSRB Reference: 2016/00281). All participants provided written informed consent.

Results

Ten index patients and 14 household contacts were included in the final analysis (Figure S3). There were seven households with one contact, two households with two contacts, and one household with three contacts. Median ages of index patients and household contacts were 59.5 and 50 years, respectively. Index patients had a higher Charlson comorbidity index, and more antibiotics and healthcare exposures compared with household contacts (Table S1). Five households completed 12 months of follow-up, two households completed 10 months, two households completed 8 months, and one household completed 6 months. The index patients and family members provided an average of 11.4 (range 8–13) and 10.7 (range 6–13) stool samples, respectively.

Index patients were colonized with bla_OXA-48-like (n = 4), bla_KPC-2 (n = 3), bla_IMP (n = 2), and bla_NDM-1 (n = 1), distributed among divergent species of Enterobacteriaceae. After a cumulative follow-up time of 9.0 years, three family members (21.4%, 3/14) acquired four different types of CPE in the community (Table 1). Of these four acquisitions, two (14.3%, 2/14) (bla_OXA-48-like and bla_KPC-2) met the definition of transmission events (hazard rate, 0.22/year; 95% CI 0.06–0.89).

Table 1.

Result of CPE acquisition and transmission among family members

Family unit (n = 10)	Index CPE genotype	Index CPE species	Family members (n = 14)	Relationship to index patient	CPE acquisition	Clinical transmission event	Genomic transmission event	Shares bed with index	Charlson score	Travel history	Hospitalization history	Antibiotics exposure
502	IMP	E. coli	1	Wife	IMI-1 E. cloacae	No	No	Yes	2	Yes	No	No
504	OXA-48-like	E. coli K. pneumoniae	3	Wife	No			No	6	Yes	Yes	No
				Daughter	No			No	0	No	Yes	Yes
				Son-in-law	No			No	0	Yes	No	No
506	IMP	E. coli	1	Husband	No			Yes	7	Yes	No	No
507	OXA-48-like	E. coli	2	Wife	No			Yes	2	Yes	Yes	Yes
				Son	No			No	0	No	Yes	Yes
509	KPC-2	K. pneumoniae	1	Father	KPC-2 genea	Yes	Possible	No	3	No	No	No
511	KPC-2	E. coli K. oxytoca E. aerogenes	2	Mother	No			No	1	Yes	No	Yes
				Father	No			No	0	Yes	No	Yes
525	NDM-1	E. coli K. pneumoniae	1	Wife	No			Yes	0	Yes	No	No
528	KPC-2	E. cloacae	1	Wife	No			No	1	Yes	No	No
1664	OXA-48-like	E. coli	1	Daughter	No			No	0	No	No	No
1665	OXA-48-like	K. pneumoniae	1	Wife	OXA-48 K. pneumoniae	Yes	Possible	Yes	0	Yes	No	Yes
					IMI-1 E. cloacae	No	No		1

a

KPC-2 gene detected from the metagenomic analysis.

In the first of the two families with CPE transmission (family unit 1665), 14 OXA-48 Klebsiella pneumoniae (S0–S9) and 13 OXA-48 Escherichia coli (S1–S9) were detected in the index patient. The family member who acquired CPE during the follow-up period was negative for CPE in stool samples collected at S2, S3 and S6 but was positive for OXA-48 K. pneumoniae at S7. All of the K. pneumoniae strains in this family were ST307. The SNP distance of 136.25 from WGS indicated a close relationship between these strains. At S7, we also detected IMI-1-producing Enterobacter cloacae in the family member which was not present in the index patient. No CPE was detected in this family member in the subsequent two stool samples (S8 and S9).

In the other family with CPE transmission (family unit 509), only the baseline sample (S0) from a total of 10 stool samples collected from the index was positive for KPC-2-producing K. pneumoniae by culture method. However, stool microbiome analysis revealed evidence for presence of bla_KPC-2 and plasmid contigs similar to those of sample S0 in samples S1–S12. The associated family member who acquired CPE during the follow-up period was negative for CPE by culture method throughout the follow-up. However, stool microbiome analysis revealed evidence for presence of bla_KPC-2 and plasmid contigs in sample S3 but not in samples S0–S2 and S4–S10.

Transmission from an index patient to a household member occurred at a rate of 0.005/colonized person/day (95% CI 0.0005–0.05) and background transmission due to other sources at 0.001/day (95% CI 0.0001–0.01). Decolonization occurred at 0.03/day (95% CI 0.01–0.06) for index patients and 0.04/day (95% CI 0.002–1.2) for household members (Table S2). The mean time that an index patient remains a risk for the household members was estimated to be 32 days (95% CI 15–66). The probability of CPE transmission from an index patient to a household contact was 10% (95% CI 4%–26%). These estimates are independent of household size and the number of colonized individuals within a household.

Eight of the 10 households agreed to environmental sampling. A total of 472 samples were collected from door handles, living room chairs, beds, toilet flush handles, kitchen water taps and chopping boards. All samples were negative for CPE. Three cats were recruited from two households; all 13 stool samples from these animals were negative for CPE.

Discussion

Previously, the introduction of CPE into the community has been shown to occur via food sources,⁷ travel to endemic countries,⁸ and acquisition from companion animals.⁹ However, household transmission of CPE has not been studied systematically except for occasional case reports.^10,11 As a first, in this cohort study, we found that the probability of CPE transmission from an index patient to a household contact was 10% and the colonization of CPE among household members was transient.

The household transmission rate of CPE was far lower than the recently reported household transmission rate of ESBL-producing bacteria in the Netherlands, which estimated a probability of transmission of 67% (95% CI 38%–88%)¹² when comparing point estimates. The conclusion was the same when taking into account the uncertainty associated with the sample. This could be due to underlying differences in the community prevalence and transmission dynamics of ESBL-producing bacteria and CPE. The main strength of this study is the use of a combination of genomics and clinical epidemiology to determine the presence of household transmission. Although it is impossible to identify transmission events with certainty, in two cases where exposed household members screened negative for CPE before the index patient was discharged from hospital and acquired CPE only after exposure to the index patient, the genomic analysis was consistent with household transmission (of bla_KPC-2 and bla_OXA-48) and provides a parsimonious explanation for these observations.

This study has several limitations. First, we acknowledge that the small sample size of the study limited detailed analyses of the clinical factors associated with transmission. However, we believe this pilot study will inform sample size calculation for future investigations. For example, based on the estimated transmission proportion of 10%, desired precision of 0.05, and 95% CI, and taking into account the drop-out rate, a total number of 150 CPE-positive index cases would need to be recruited. Second, the results should be interpreted in the context of a high-income country with a low prevalence of community-onset CPE and with an advanced sanitation system. Third, the environmental samples did not undergo an enrichment step prior to culture, which may have reduced the sensitivity of CPE identification. Finally, there were two occasions where IMI-1-producing E. cloacae was identified in a family member but not in the associated index patient. The two E. cloacae were not related by genome sequencing, and had distinct MLST sequence types and resistance gene profiles, suggesting that these were independently acquired in these two family members. However, we were unable to confirm their sources.

Acknowledgements

The investigators thank the infection prevention and control units of Tan Tock Seng Hospital and Singapore General Hospital for facilitating the conduct of this study.

Funding

The was supported by the Singapore Biomedical Research Council-Economic Development Board Industry Alignment Fund (IAF311018), National Medical Research Clinician Scientist Individual Research Grant (CS-IRG) (CIRG18nov-0034), National Medical Research Council Collaborative Grant (NMRC CGAug16C005), and National Medical Research Council Clinician Scientist Award (NMRC/CSA-INV/0002/2016 and MOH-CSAINV18nov-0004).

Transparency declarations

None to declare.

Supplementary data

Additional Methods, Tables S1 and S2, and Figures S1 to S3 are available as Supplementary data at JAC Online.

Author notes

Niranjan Nagarajan, Swaine L Chen and Timothy Barkham Contributed equally.