Controlling intestinal colonization of high-risk haematology patients with ESBL-producing Enterobacteriaceae: a randomized, placebo-controlled, multicentre, Phase II trial (CLEAR)

Abstract

Objectives

We assessed the efficacy and safety of an oral antimicrobial regimen for short- and long-term intestinal eradication of ESBL-producing Escherichia coli and Klebsiella pneumoniae (ESBL-EC/KP) in immunocompromised patients.

Methods

We performed a randomized (2:1), double-blind multicentre Phase II study in four haematology–oncology departments. Patients colonized with ESBL-EC/KP received a 7 day antimicrobial regimen of oral colistin (2 × 10⁶ IU 4×/day), gentamicin (80 mg 4×/day) and fosfomycin (three administrations of 3 g every 72 h), or placebo. Faecal, throat and urine specimens were collected on day 0, 6 ± 2, 11 ± 2, 28 ± 4 and 42 ± 4 after treatment initiation, and the quantitative burden of ESBL-EC/KP, resistance genes and changes in intestinal microbiota were analysed. Clinicaltrials.gov: NCT01931592.

Results

As the manufacture of colistin powder was suspended worldwide, the study was terminated prematurely. Overall, 29 (18 verum/11 placebo) out of 47 patients were enrolled. The short-term intestinal eradication was marginal at day 6 (verum group 15/18, 83.3% versus placebo 2/11, 18.2%; relative risk 4.58, 95% CI 1.29–16.33; Fisher’s exact test P = 0.001) and not evident at later timepoints. Quantitative analysis showed a significant decrease of intestinal ESBL-EC/KP burden on day 6. Sustained intestinal eradication (day 28 + 42) was not achieved (verum, 38.9% versus placebo, 27.3%; P = 0.299). In the verum group, mcr-1 genes were detected in two faecal samples collected after treatment. Microbiome analysis showed a significant decrease in alpha diversity and a shift in beta diversity.

Conclusions

In this prematurely terminated study of a 7 day oral antimicrobial eradication regimen, short-term ESBL-EC/KP suppression was marginal, while an altered intestinal microbiota composition was clearly apparent.

Introduction

Worldwide, the prevalence of colonization and infection with ESBL-producing Enterobacteriaceae (ESBL-E) is increasing.^1–3 Particularly in immunocompromised hosts, bloodstream infections (BSIs) caused by these organisms represent a major clinical challenge, given the limited antimicrobial treatment options and poor outcome.^4,5 Colonization by ESBL-E has been shown to predispose to BSIs caused by these organisms.⁶ Thus, in order to decrease infection rates, antimicrobial decolonization strategies have been suggested. However, the available studies differ considerably in design, choice of antimicrobial regimen, dosage and duration of therapy.^7–11 The only randomized (1:1), double-blind, placebo-controlled study assessing the clinical efficacy of oral colistin and neomycin for 10 days (plus nitrofurantoin in the case of ESBL-E bacteriuria) for the eradication of ESBL-E from the intestine of 54 patients showed only temporary success.⁷ Similar results were reported from a decolonization study in rectal carriers of carbapenem-resistant Klebsiella pneumoniae who received topical gel application orally and an oral solution of gentamicin and colistin.⁸ Furthermore, a meta-analysis showed a protective effect on colonization and/or infection by Gram-negative bacteria resistant to third-generation cephalosporins (in most cases ESBL-E) for patients receiving selective oropharyngeal decontamination (SOD) or selective digestive decontamination (SDD) (OR 0.33, 95% CI 0.20–0.52).¹² However, these strategies imply that a large number of patients may experience unnecessary exposure to SOD or SDD, while very few will benefit. Such extended antimicrobial exposure potentially leads to a perturbation of the intestinal microbiota, selection of antimicrobial-resistant bacteria and the exchange of antimicrobial resistance genes through horizontal gene transfer (HGT) between bacterial species.¹³ Specifically in haematological and oncological patients, these factors have been shown to impact outcome.^14,15 A recently published European guideline on decolonization of MDR Gram-negative bacteria carriers concluded that current evidence is too limited to recommend eradication and suggests prospective clinical studies in immunocompromised patients.¹⁶

Since the risk of subsequent BSI in colonized patients is highest during periods of immunosuppression and neutropenia, we assessed the clinical safety and efficacy of oral fosfomycin, colistin and gentamicin in the short-term and long-term eradication of ESBL-E from the gastrointestinal tract in this specific high-risk setting.

Patients and methods

Study design and endpoints

This randomized, double-blind, placebo-controlled, multicentre Phase II trial was conducted in the haematology/oncology departments of four German university hospitals. As there were previously published data available on the treatment components, a 2:1 allocation (verum:placebo) was used.⁸ Following written informed consent, randomization was carried out (Supplementary Methods, Text S1, available as Supplementary data at JAC Online). In two parallel groups, an antimicrobial eradication regimen of 7 days oral colistin (2 × 10⁶ IU 4×/day), gentamicin (80 mg 4×/day) and three administrations of fosfomycin (3 g every 72 h) was compared with placebo with regard to the eradication of ESBL-producing Escherichia coli or K. pneumoniae (ESBL-EC/KP) in high-risk patients. See Supplementary Methods Text S2 on choice of study treatment and Text S3 on study medication details.

Patients were eligible for study inclusion if faecal screening within 14 days prior to enrolment revealed colonization with ESBL-EC/KP, and if they were considered severely immunocompromised (haematopoietic stem cell transplantation or chemotherapy with expected neutropenia ≥3 days, solid organ transplantation or administration of high-dose corticosteroids or other immunosuppressants, or expected neutropenia of at least 7 days due to an underlying condition, including functional neutropenia). Exclusion criteria included current or already scheduled administration of ESBL-E active antimicrobial treatment, and liver or kidney dysfunction. See Supplementary Methods Text S4 for a comprehensive list of inclusion and exclusion criteria.

Short-term intestinal eradication, defined as a faecal sample negative for ESBL-EC/KP on both days 6 ± 2 and 11 ± 2, was the primary endpoint. Secondary endpoints were long-term intestinal eradication on day 28 ± 4 and 42 ± 4, short- and long-term non-intestinal eradication defined as urine and throat samples negative for ESBL-EC/KP in patients with initially positive samples, and detection of non-ESBL MDR bacteria in the intestine including VRE, carbapenem-resistant Gram-negative bacteria (except Proteus and Serratia) and colistin-resistant Gram-negative bacteria. Resistance genes in the intestine were assessed on day 11 ± 2 and 42 ± 4 and compared with their presence at baseline. Furthermore, the quantitative burden of ESBL-EC/KP during follow-up was analysed. As exploratory endpoints, dynamics in the intestinal microbiome composition were assessed by 16S rRNA gene sequencing. The original study protocol is available as Supplementary data (File 2).

Sample and data collection

Faecal samples (stool samples, alternatively rectal swabs), throat swabs and urine specimens were collected on day 0 (baseline), day 6 ± 2, 11 ± 2, 28 ± 4 and 42 ± 4. For determination of blood chemistry, samples were drawn on day 0 and 6 ± 2. Throughout the study period, adverse events (AEs), severe adverse events (SAEs) and AE-related study drug discontinuations were assessed. Documentation included demographics, underlying disease, cytotoxic and immunosuppressive therapies, antimicrobial administration, neutropenia and occurrence of BSIs with ESBL-EC/KP. We assessed whether the administered antimicrobial drugs were potentially active against ESBL-E and individually active according to the susceptibility profile of the respective baseline ESBL-EC/KP isolate.

Microbiological methods

Samples were processed locally. For detection of ESBL-EC/KP, collected samples were plated on selective media (chromID^® ESBL, bioMérieux, Nürtingen, Germany), incubated for 18–24 h and colonies identified to species level by MALDI-TOF. Susceptibility testing was performed using Vitek2 (bioMérieux) including colistin resistance; production of ESBL was confirmed by CLSI combination disc test as previously described.³ In order to increase the diagnostic yield, an aliquot of a 1:10 dilution of stool or undiluted rectal swab medium was suspended in a selective MacConkey enrichment broth for 18–24 h and then processed as described above.^17,18

For quantitative analysis of ESBL-EC/KP, 10-fold serial dilutions (100 μL each) from stool samples were plated on MacConkey agar and selective media (chromID^® ESBL) and incubated for 18–24 h. Afterwards, colonies on the plate with the highest dilution showing bacterial growth were counted and bacterial concentrations noted as cfu/g of stool. If ESBL-EC/KP were only detected after enrichment, the bacterial concentration was arbitrarily set as 50 cfu/g given the lowest detection level of 100 cfu/g for direct culture. For detection of VRE, 10 μL of the 1:10 solution described above was plated on selective medium (chromID^® VRE, bioMérieux) and incubated for 48 h. Vancomycin resistance as determined by Vitek2 was confirmed by Etest in a second step.

Microbiome analysis

For microbiome analysis, an aliquot from stool samples was transferred to a storage medium using a swab (COPAN FecalSwab™, Brescia, Italy) and frozen at −80°C. After trial completion, samples were subjected to DNA extraction using the FastDNA Spin Kit for Soil (MP Biomedicals, Solon, OH, USA). Afterwards, 16S rRNA amplicon sequencing of the V3–V4 region was performed as outlined in the Illumina 16S Sample Preparation Guide and sequenced on the MiSeq platform (Illumina, San Diego, CA, USA).¹⁹ Sequencing data were processed using the DADA2 pipeline and QIIME version 2.^20,21 Taxonomic classification was performed by a Naive Bayes classifier (sklearn).²²

After rarefaction at a sequencing depth of 7500 sequences per sample, alpha diversity (diversity within one sample) and beta diversity (diversity between samples) metrics were calculated using the R package phyloseq.^23,24 Continuous data were presented as the mean (±SD) and/or median (range) and tested with appropriate statistical analysis. Absolute and relative frequencies were given for categorical data (i.e. dominations) and tested using the χ² test. Domination was defined as a single bacterial taxon comprising at least 30% of sequences and being the most abundant taxon.¹⁴ The beta diversity, in this case the generalized UniFrac distances between the samples, was visualized using principal coordinate analysis (PCoA), and differentially abundant taxa were identified using Linear Discriminant Analysis (LDA) Effective Size (LEfSe).²⁵

Antimicrobial resistance genes (ARGs)

The presence of 16 ARGs was analysed in stool samples obtained at days 0, 11 and 42 in patients with complete sample collections and without treatment discontinuation.

For ARG detection, custom multiplex TaqMan assays were developed as described previously.²⁶ For details on primer design see Supplementary Methods, Text S5 and Table S1.

Statistical analysis

Statistical analysis was carried out using SPSS software (IBM SPSS Statistics, version 23; IBM, Armonk, NY, USA). Based on prior data, we expected colonization with ESBL-EC/KP in 84% of patients without treatment,⁶ and in 40% in the eradication arm.⁸ Aiming at 80% power and accounting for dropouts and stratification, sample size calculations resulted in a total of 47 subjects (31 in the verum, and 16 in the placebo arm) to be randomized (Supplementary Methods Text S6). The ITT population included all randomized subjects. The PP population involved only patients with complete treatment administration (Supplementary Methods Text S7). Furthermore, a subanalysis was conducted including only those samples obtained without prior exposure to individually active antimicrobial drugs according to susceptibility testing of the baseline ESBL-EC/KP isolate [microbiologically modified ITT (micromITT) analysis]. Patients with exposure to individually active antimicrobials prior to day 6 sampling were excluded.

The primary endpoint was evaluated by means of the two-sided Mantel–Haenszel test stratified by study centre including calculation of the common OR with 95% CI. For endpoint assessment, missing samples were counted as positive for ESBL-EC/KP. Distribution of data between groups was described using count, percentage, mean, median and IQR, and compared using Fisher’s exact test, Wilcoxon signed-rank test, Kruskal–Wallis test or Skillings–Mack test (incomplete data sets, performed in R), as appropriate. Relative risk (RR) with the corresponding 95% CI were calculated.

Ethics

The study was conducted in accordance with the Declaration of Helsinki in the version of October 2008 (59th WMA General Assembly, Seoul), the internationally recognized Good Clinical Practice Guidelines (ICH-GCP) and German Drug Law (AMG). It was approved by the competent ethics committees (Cologne-ID #13-271) and the Federal Institute for Drugs and Medical Devices (#4039623). The study has been registered at Clinicaltrials.gov (NCT01931592).

Results

Patient characteristics

Between April 2014 and February 2016, 29 out of 47 patients were enrolled into the study, 18 in the verum and 11 in the placebo arm (Figure 1). Since the manufacture of colistin powder was suspended worldwide, we were forced to terminate the trial prematurely. All 29 patients were evaluable in the ITT analysis. Demographic characteristics, exposure to chemotherapy and immunosuppression, and median duration of neutropenia did not differ significantly between the groups (Table 1). At baseline, faecal samples of 27 patients tested positive for ESBL-producing E. coli (ESBL-EC) and two for ESBL-producing K. pneumoniae (ESBL-KP). Twenty-one patients (13 verum/8 placebo) received the complete treatment (72%, PP population). Four patients from the verum group discontinued the intake of study medication before treatment day 4 owing to gastrointestinal disorders. In one patient receiving placebo, the physician decided to end treatment on day 6 owing to increased creatinine levels. For three patients (one verum/two placebo), minor deviations related to study drug administration such as one missing dose or delayed intake were recorded. For evaluation of endpoints, samples from one patient on day 11 (verum) and three patients on day 28 and 42 (all verum) were missing. Rectal swabs were obtained instead of stool samples in five visits (2 verum patients/1 placebo).

Table 1.

Demographic and clinical characteristics of included patients

Characteristics	Verum group (n = 18)	Placebo group (n = 11)	P value
Age, years, median (range; IQR)	52 (30–73; 39–63)	53 (28–64; 48–60)	0.840a
Gender (male/female)	13/5	8/3	1.000b
BMI, kg/m2, median (range; IQR)	25.5 (21.6–32.6; 22.7–28.8)	25.1 (19.7–40.3; 23.1–29.7)	0.928a
Underlying conditions, n (%)			0.387b
acute leukaemia	3 (16.7)	2 (18.2)
Hodgkin’s lymphoma	3 (16.7)	0
myeloma	1 (5.6)	3 (27.3)
non-Hodgkin’s lymphoma	9 (50)	4 (36.4)
solid tumour	2 (11.1)	2 (18.2)
ECOG score at randomization (%)c			0.686a
0	7 (38.9)	5 (45.5)
1	8 (44.4)	4 (36.4)
2	3 (16.7)	1 (9.1)
3	0	1 (9.1)
Inclusion group (%)			0.349b
immunosuppressive therapy	0	1 (9.1)
stem cell transplantation	3 (16.7)	3 (27.3)
chemotherapy	15 (83.3)	7 (63.6)
Exposure to chemotherapy, n (%)	16 (88.9)	9 (81.8)	0.622b
Exposure to immunosuppressants, n (%)	13 (72.2)	8 (72.7)	1.000b
Duration of neutropeniad, days, median (range; IQR)	7.5 (0–25; 4–11)	8 (0–29; 3–14)	0.804a

ECOG, European Cooperative Oncology Group.

a

Mann–Whitney U test.

b

Fisher’s exact test.

c

No ECOGs of 4 or 5 were reported.

d

Neutropenia defined as <500 neutrophils/mL or white blood count <1000 leucocytes/mL.

Figure 1.

CONSORT flow diagram. MicromITT, microbiologically modified ITT including only those samples without prior exposure to antimicrobials tested as susceptible in the resistance testing of their specific baseline ESBL-EC/KP isolate. Patients with such exposure prior to day 6 sampling were excluded; PP analysis includes only patients without treatment discontinuation or incomplete administration of medication.

Primary and secondary endpoints

The primary endpoint of short-term intestinal eradication (day 6 and 11) differed at marginal significance between the two groups, with 11/18 (61.1%) patients in the verum group tested negative versus 2/11 (18.2%) in the placebo group (P = 0.043 as per protocol, Mantel–Haenszel common OR estimate 0.06; 95% CI, 0.01–0.70; RR 3.36, 95% CI 0.91–12.42; Fisher’s exact test P = 0.052). Short-term eradication on day 6 only was significantly higher in the verum group, with an RR of 4.58 (95% CI 1.29–16.33).

Rates of long-term (day 28 and 42) intestinal eradication were not statistically significant between groups [verum, 7/18 (38.9%) versus placebo, 3/11 (27.3%); Table 2]. In the PP analysis, 10/13 (76.9%) patients in the verum and 1/8 patients (12.5%) patients in the placebo group were tested negative for ESBL-EC/KP during the short-term period (P = 0.020); long-term eradication rates did not differ significantly between groups. Endpoints of non-intestinal eradication were not calculated, due to too small incidences: ESBL-EC/KP was detected in urine in only two patients in the verum group and four in the placebo group. None of the patients displayed throat colonization at baseline. During follow-up, three verum patients were colonized in their urine and two verum patients in their throats (Table S2). A carbapenem-resistant KP was detected in one placebo patient on day 6, in another placebo patient on day 28 and in one in the verum group on day 28 only. The presence of VRE remained nearly constant during the observation period (Table S3). Colistin-resistant Gram-negative bacteria were not isolated. On day 6, quantitative analysis showed a significant decrease of intestinal ESBL-EC/KP burden in the verum compared with the placebo group (verum, mean 6.14 × 10⁷ cfu/g versus placebo, mean 2.04 × 10⁸ cfu/g; P = 0.005; Figure 2). All quantitative results are shown in Table S4.

Table 2.

Primary and secondary endpoints regarding intestinal eradication of ESBL-EC/KP

Endpoints	Verum group (n = 18)	Placebo group (n = 11)	RR (95% CI)	P value
Short-term intestinal eradication
day 6/day 11 (primary endpoint)a	11 (61.1)	2 (18.2)	3.36 (0.91–12.42)	0.043b; 0.052c
day 6	15 (83.3)	2 (18.2)	4.58 (1.29–16.33)	0.001c
day 11a	11 (61.1)	4 (36.4)	1.68 (0.71–3.99)	0.317c
Long-term intestinal eradication
day 28a	7 (38.9)	3 (27.3)	1.43 (0.46–4.39)	0.299c
day 42a	7 (38.9)	3 (27.3)	1.43 (0.46–4.39)	0.299c

RR, relative risk comparing the verum with the placebo group.

a

Samples from one patient on day 11 (verum), and three patients on day 28 and 42 (all verum) were missing and counted as positive for these endpoints.

b

Two-sided Mantel–Haenszel test as per protocol.

c

Fisher’s exact test.

Figure 2.

Intestinal ESBL-EC/KP burden measured as cfu/g stool at different timepoints. Samples within each group were compared with the baseline (day 0) samples using the Wilcoxon signed-rank test. *P ≤ 0.05; **P < 0.01; NS, not significant (P > 0.05). Furthermore, differences within groups were compared using Skillings–Mack test (complete cases only, not shown in the figure): placebo, n = 10, P = 0.231; verum, n = 11, P = 0.036. The lower whisker extends to the smallest observation greater than or equal to –1.5×IQR from the 25th percentile; the box extends from the 25th percentile to the 75th percentile and is intersected with a line at the median; the upper whisker extends from the largest observation less than or equal to the 75th percentile +1.5×IQR.

AEs, study discontinuations and laboratory results

In total, 200 AEs in 17 patients compared with 118 AEs in 11 patients were reported in the verum and placebo group, respectively (P = 1.000). Intensity of AEs did not differ significantly between groups (P = 0.149), with nearly all AEs being rated Common Toxicity Criteria (CTC) Grade 1–3 except one AE in the verum group with CTC Grade 5. Seventeen AEs in eight patients in the verum group, and 12 AEs in six patients in the placebo group, were rated as probably/likely or possibly related to the study drug (P = 0.115). Gastrointestinal disorders were the most frequent AEs. See Tables S5, S6 and S7 for further details on AE frequencies. There were four (22.2%) compared with one (9.1%) treatment discontinuations in the verum and placebo arm, respectively (P = 0.622). Another patient in the verum arm died on day 16 due to pneumonia and sepsis, constituting the only SAE in this study. No ESBL-E BSI occurred during the study period.

Blood chemistry results were compared between baseline and day 6. Laboratory changes were largely similar between groups (Table S8).

Concomitant antimicrobial exposure

Nearly all patients in both arms received concomitant antimicrobial agents during the study period (17/18 verum, 10/11 placebo), in most cases due to febrile neutropenia. A significant difference was observed only for aminoglycosides, with a higher exposure in the verum group (P = 0.036). The exposure to potentially and individually ESBL-E active antimicrobials was similar in both groups (Table 3).

Table 3.

Exposure to concomitant antimicrobial drugs

Characteristic	Verum group (n = 18)	Placebo group (n = 11)	P value
Number of patients with any antimicrobial drugs, n (%)	17 (94.4)	10 (90.9)	1.000a
Antimicrobial drugs by class, n (%)
aminoglycosides	6 (33.3)	0 (0.0)	0.058a
sulphonamides	16 (88.9)	9 (81.8)	0.622a
carbapenems	8 (44.4)	5 (45.5)	1.000a
β-lactam/β-lactamase inhibitor combinations	4 (22.2)	4 (36.4)	0.433a
fluoroquinolones	5 (27.8)	4 (36.4)	0.694a
fosfomycin	1 (5.6)	0 (0.0)	1.000a
third-generation cephalosporins	7 (38.9)	1 (9.1)	0.110a
other antimicrobials	5 (27.8)	1 (9.1)	0.362a
Duration, days, mean/median
aminoglycosides	1.2/0	0/0	0.036b
sulphonamides	11.8/14.5	8.5/9	0.241b
carbapenems	3.9/0	3.9/0	0.921b
β-lactam/β-lactamase inhibitor combinations	1.6/0	3.4/0	0.376b
fluoroquinolones	4.3/0	3.5/0	0.848b
fosfomycin	0.4/0	0/0	0.434b
third-generation cephalosporins	1.0/0	0.2/0	0.081b
other antimicrobial drugs	2.1/0	2.8/0	0.310b
Patients with potentially ESBL-E active antimicrobials, n (%)	16 (88.9)	10 (90.9)	1.000a
Patients with individually ESBL-EC/KP active antimicrobials, n (%)	11 (61.1)	7 (63.6)	1.000a
Duration, days, mean/median
potentially ESBL-E active antimicrobial drug	18.9/17.5	15.8/13	0.300b
individually ESBL-EC/KP active antimicrobial drug	4.9/2.5	5.3/8	0.595b

ESBL-E, ESBL-producing Enterobacteriaceae; ESBL-EC/KP, ESBL-producing Escherichia coli or Klebsiella pneumoniae.

a

Fisher’s exact test.

b

Mann–Whitney U test.

micromITT analysis

The primary endpoint of short-term intestinal eradication also differed significantly between the two groups, with 10/13 (76.9%) patients in the verum group being tested negative versus 0/6 (0%) in the placebo group (P = 0.042). Again, long-term intestinal eradication on day 28 and 42 did not differ significantly between the two groups (Table S9).

Microbiome analysis

Overall, microbiome analysis was performed for 78 faecal samples from verum and 54 from placebo patients. The 12 remaining samples (11 verum/1 placebo) were either missing or collected as rectal swabs; one day 28 verum sample failed sequencing. During and shortly after the intervention, there was a more pronounced decrease in alpha diversity (Shannon Index and Faith’s Phylogenetic Diversity) in the verum group as compared with the placebo group (Table S10). In the verum group, alpha diversity was significantly lower in samples from day 6 and day 11 as compared with baseline (Figure 3). This temporary decrease was more pronounced when including only samples obtained without prior exposure to broad-spectrum antimicrobial treatment including carbapenems, third-generation cephalosporins, β-lactam/β-lactamase inhibitor combinations, fluoroquinolones and aminoglycosides (Table S11 and Figure S1). Furthermore, we observed longitudinal changes with respect to the composition in both groups (Figure S2). Observed rates of intestinal domination by Enterococcus spp. (genus level) or Enterobacteriaceae (family level) were similar in both arms (Table S12). All cases of enterococcal domination in the placebo group were detected in patients receiving broad-spectrum antimicrobial treatment prior to the respective sampling. LEfSe revealed a higher abundance of Bacteroides spp. (genus level) and Bacteroidaceae (family level) in verum samples compared with placebo most markedly present on day 6 (Figure S3). On day 6, samples from 13/18 (72.2%) patients in the verum group as compared with samples from 1/11 (9.1%) in the placebo group displayed intestinal domination by Bacteroides spp. (P = 0.001; Table S12). A significant shift of the microbial composition between the placebo and verum group was confirmed by analysis of similarity (ANOSIM) based upon generalized UniFrac (alpha = 0.5) distances measurements for day 6 (ANOSIM R²=0.477, P = 0.001) (Figure S4).

Figure 3.

Alpha diversity [Shannon Index (a) and Faith’s Phylogenetic Diversity (b)] in verum and placebo samples 0, 6, 11, 28 and 42 days after treatment. Samples within each group were compared with the baseline (day 0) samples using the Wilcoxon signed-rank test. *P ≤ 0.05; **P < 0.01; ***P < 0.001; NS, not significant (P > 0.05). Furthermore, differences within groups were compared using Skillings–Mack test (complete cases only, not shown in the figure): (a) Shannon Index: placebo, n = 10, P = 0.0842; verum, n = 12, P = 0.0180; (b) Faith’s Phylogenetic Diversity: placebo, n = 10, P = 0.3415; verum, n = 12, P = 0.0049. See Figure 2 legend for an explanation of the convention for the box and whisker diagrams.

Resistance genes

In total, 38 faecal samples from 13 verum patients and 29 samples from 10 placebo patients were available for assessment of antimicrobial resistance genes (Figure S5). None of the tested carbapenemase-encoding genes was detected in the sample collection. Fluoroquinolone resistance genes (qnrB) were more frequently detected in verum patients, in particular during follow-up (one patient at baseline, two on day 11 and four on day 42), than in placebo patients (1 patient at baseline). Furthermore, there were two patients in the verum group with samples on day 42 being positive for the mcr-1 gene, which mediates colistin resistance. In these two cases, the PCR product was subjected to Sanger sequencing. Basic Local Alignment Search Tool (BLAST) search against the Comprehensive Antibiotic Resistance Database (CARD) confirmed the identity of the sequenced amplicons as the mcr-1 gene (Figure S6).²⁷ The collected ESBL-EC/KP of these two patients were PCR negative for mcr-1 and also for mcr-2, mcr-3, mcr-4 and mcr-5. Of note, neither of these two patients had received colistin apart from the study treatment.

Discussion

Patients receiving the eradication treatment combination were more likely to test negative for ESBL-EC/KP until day 11 (61% versus 18%) with a clearly significant difference for day 6 only (RR 4.58, 95% CI 1.29–16.33). However, this effect was no longer detectable on day 28 and 42, even if the analysis was restricted to patients who had not received ESBL-EC/KP active concomitant antimicrobial treatment based on the individual susceptibility profile of the respective isolate. The study medication was well tolerated.

Our results are consistent with those of two previous randomized trials, which both reported successful short-term eradication only.^7,8 We hypothesize that patients never truly cleared their intestinal ESBL-EC/KP, but that bacterial density was intermittently reduced to below the detection limit of culture methods. It is unknown whether the risk of translocation of ESBL-EC/KP strains into the bloodstream is reduced during this period of low-density colonization. Considering the fact that the human gut microbiota consists of up to 10¹⁴ bacteria, a decrease seems likely.²⁸ However, a large randomized trial (including several thousand patients, based on published frequencies of ESBL-E BSI) would be needed to prove the efficacy of our oral eradication regimen for the prevention of BSI with ESBL-E during neutropenia.^6,29

A rapidly growing body of evidence indicates an important role for the gut microbiota in immune homeostasis and intestinal colonization resistance.³⁰ In the verum group, disturbance of the intestinal microbiota with a decrease in alpha diversity and a shift in beta diversity was observed on day 6 and 11. Previous studies have shown a correlation between loss of diversity, emergence of dominance of pathogenic bacteria and the subsequent risk of invasive infections with these pathogens.^14,31,32 Therefore, our findings might be associated with an increased translocation and infection risk due to changes in the composition of the gut microbiota.³³ It may thus be speculated that antimicrobial drugs administered for the purpose of decolonization may indeed prevent clinically relevant infections during the time of exposure, but eventually impact negatively on patient outcome.

Further selection of antimicrobial resistance is another aspect that deserves specific attention in this context. Selective culture and susceptibility testing revealed comparable frequencies of newly detected VRE in both groups, and no detectable emergence of colistin-resistant isolates. However, assessment of antimicrobial resistance genes showed the mcr-1 gene encoding colistin resistance in two verum faecal samples from day 42. It is possible that bacteria harbouring this gene multiplied under colistin exposure, but remained under the threshold level of culture detection methods. In previous studies, assessing SDD with colistin, the emergence of colistin resistance has been reported.^15,34,35 However, since mcr-1 can be transmitted via HGT, it is also likely that the gut microbiota themselves might have become a reservoir for this antimicrobial resistance gene. This demonstrates the potential harm of empirical or prophylactic colistin use and questions the adequacy of intestinal antimicrobial eradication regimens in general.

More recently, faecal microbiota transfer (FMT) has been employed as an alternative eradication approach for ESBL-E.^36–41 Findings so far are promising, but clearly require validation in randomized controlled trials.

Our study has some limitations: the trial was discontinued prematurely due to an unforeseeable cessation of the production of colistin powder by the manufacturer. Up until then, enrolment of patients was more challenging than expected. Eligible patients refused participation due to difficulties in understanding the relevance of their intestinal colonization and reluctance to take the necessary amount of study drugs. With this rather small sample size, favourable and unfavourable effects of the eradication regimen may have remained undetected. In addition, about two-thirds of patients in both groups received ESBL-E active antimicrobials, probably affecting the eradication endpoints. Since the completion of our study, disadvantages of using the Etest for detection of colistin resistance, and Vitek2 for vancomycin resistance, have become apparent; this is a limitation we could not be aware of at the time of the study. Regarding our main endpoints, however, we have tried to maximize the diagnostic yield for ESBL-EC/KP by use of enrichment broths in contrast to previous studies.¹⁷

In conclusion, a 7 day antimicrobial treatment regimen showed a short-term suppression of intestinal ESBL-EC/KP carriage only. With regard to antimicrobial-associated disturbances of the gut microbiota and increased rates of antimicrobial resistance genes that may potentially result in limited treatment options, we would not recommend this prophylactic antimicrobial treatment. Further studies are needed to elucidate the impact of increased selection pressure and microbiota disturbances on the infection risk and the accumulation of antimicrobial resistance genes in this high-risk patient population.

Acknowledgements

We thank the pharmacists Marija Tubic-Grozdanis and Kerstin Maiwald for providing blinded study medications, and all study site staff involved in recruitment and documentation. Preliminary results of this study were presented at the Congress for Infectious Diseases and Tropical Medicine (KIT), Cologne, Germany 20–23 June 2018; Abstract ID: FV72.

Funding

This work was supported by the German Center for Infection Research (DZIF) (DZIF grant numbers TTU08.803, TTU08.911).

Transparency declarations

L. M. B. has received lecture honoraria from Astellas Pharma and MSD, and travel grants from 3 M and Gilead. H. R. received honoraria from MSD, Gilead, Pfizer, Accelerate and Infectopharm, and research funding from Pfizer. M. v. L. T. has received honoraria and travel support from MSD, Gilead, Janssen and Celgene, and research funding from Novartis and Gilead. H. Seifert reports research grants from Accelerate, Cubist and Novartis, and personal fees from 3 M, Astellas, Basilea, Becton Dickinson, Genentech, InfectoPharm, Roche Pharma, Gilead, Merck/MSD, Tetraphase and ThermoFisher. M. J. G. T. V. has served on the speakers’ bureau of Akademie für Infektionsmedizin, Ärztekammer Nordrhein, Astellas Pharma, Basilea, Gilead Sciences, Merck/MSD, Organobalance and Pfizer, received research funding from 3 M, Astellas Pharma, DaVolterra, Gilead Sciences, MaaT Pharma, Merck/MSD, Morphochem, Organobalance, Seres Therapeutics, Uniklinik Freiburg/Kongress und Kommunikation, and is a consultant to Alb-Fils Kliniken GmbH, Arderypharm, Astellas Pharma, Berlin Chemie, DaVolterra, MaaT Pharma and Merck/MSD. All other authors have none to declare.

Author contributions

V. D., L. M. B., A. H., H. Seifert and M. J. G. T. V. conceived and designed the study. V. D., L. M. B. and M. J. G. T. V. coordinated study conduct. W. V., D. D., P. Schafhausen and M. v. L. T. enrolled patients in the study and ensured local study conduct, A. H., H. Seifert, S. P. and H. R. performed microbiological methods and resistance testing, F. F. and A. T., supported by P. G. H. performed microbiome analysis, T. E. K., P. Slickers, R. E. and H. Slevogt performed ARG analysis, H. C. and M. H. performed the statistical analysis. V. D. and L. M. B. wrote the original draft, and all authors reviewed and edited the manuscript.

Author notes

Vassiliki Dimitriou and Lena M Biehl authors contributed equally.