![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Previous Article | Next Article 
Journal of Clinical Microbiology, February 2008, p. 560-567, Vol. 46, No. 2
0095-1137/08/$08.00+0 doi:10.1128/JCM.00832-07
Copyright © 2008, American Society for Microbiology. All Rights Reserved.
Barts and The London NHS Trust, Whitechapel, London E1 1BB, United Kingdom,1 Homerton University Hospital NHS Foundation Trust, Homerton Row, London E9 6SR, United Kingdom,2 Barts and The London NHS Trust, Queen Mary's School of Medicine and Dentistry, Wolfson Institute, Charterhouse Square, London EC1M 6BQ, United Kingdom,3 University College Hospital, Euston Road, London, United Kingdom4
Received 19 April 2007/ Returned for modification 1 October 2007/ Accepted 13 November 2007
 |
ABSTRACT |
|---|
 Top ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXREFERENCES |
|---|
Enterobacteriaceae strains resistant to three or more classes of antibiotic are endemic in many hospitals and account for a significant proportion of hospital-acquired bloodstream infections (8). Outbreaks of infection have been described for many different groups of patients (9, 10, 13, 24, 43), even extending to nonhospitalized individuals (5, 9, 15, 32).
The choice of antibiotic for empirical therapy may be an important determinant of the clinical outcome for bloodstream infections caused by gram-negative bacteria (18, 42). An increasing prevalence of multiresistant Enterobacteriaceae strains (MRE; resistant to three or more classes of antibiotic) increases the probability of MRE infection in vulnerable infants, with the potential for worse outcomes for infants associated with suboptimal empirical antibiotic treatment (21, 44). Many outbreaks of infection with antimicrobial-resistant Enterobacteriaceae strains have been described for infants nursed in neonatal intensive care units (NICUs) (4, 7, 11, 20).
High rates of the colonization of infants in special-care baby units with MRE suggest that, in some units, MRE are endemic. There is evidence from previous studies, including a pilot study in a NICU in East London (31), that a large proportion of infants nursed in intensive care units acquire colonization with MRE, but there is little information on the duration of carriage in this population. Persistent colonization with resistant Enterobacteriaceae strains has potential implications not only for colonized infants but also for their contacts (12, 22).
In this study, we have determined the frequency of and risk factors for the carriage of resistant strains by infants from two NICUs during the initial hospital stay and the persistence of resistant strains through infancy.
|
MATERIALS AND METHODS |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXREFERENCES |
|---|
Infants with suspected early-onset sepsis are treated with penicillin and gentamicin in both units. Unit 1 uses piperacillin-tazobactam with vancomycin, and unit 2 uses flucloxacillin and gentamicin for suspected late-onset sepsis.
Approval for this study was given by East London and the City Research Ethics Committee; all of the parents of participating infants gave informed consent.
Inclusion criteria. All babies born during the study period and admitted to the unit within 24 h of birth and who remained for more than 14 days were eligible for recruitment into the study.
Exclusion criteria. Babies with major congenital malformations and babies of non-English speakers for whom no advocate could be found were excluded. Those who died or who required gastrointestinal surgery before discharge from the hospital, those from whom samples were not obtained while in the hospital, and those from whom follow-up samples were unlikely to be obtained also were excluded.
Sampling. Stool samples were collected at 2 weeks postnatal age and at the time of discharge. Parents also were approached to provide samples of stool from the infants at 6 and 12 months after discharge.
Data collection. The initial clinical information included sex, gestational age, birth weight, ethnicity, maternal group B streptococcus (GBS) status (if known), the presence of a prolonged rupture of membranes (>24 h), the use of antibiotics in pregnancy and intrapartum, the use of antenatal corticosteroids, the number of births (i.e., single birth or multiple births), the mode of delivery, the use of mechanical ventilation, the use of a surfactant at birth, and the severity of early illness (clinical risk index for babies [CRIB] II score) (29). Additional data were collected each day and included the use and type of antibiotic, probable sepsis status (i.e., a positive culture from a normally sterile sample such as blood culture or cerebrospinal fluid), the type of feeds, the duration of ventilation via a tracheal tube, the use of a orogastric/nasogastric tube, the use of ranitidine, whether or not the infant had continued oxygen dependence at 28 weeks postnatal age and 36 weeks postmenstrual age, and the length of stay.
At the time of collecting the 1-year-postdischarge stool sample, information about hospitalizations and antibiotic use during infancy was requested from the mother using a standardized questionnaire.
Laboratory methods. Samples were processed in the routine microbiology laboratory at Barts and the London NHS Trust. Stool was added to a cryopreservative (brain heart infusion broth containing 10% glycerol, 0.0001% resazurin, and 0.02% dithiothreitol) and stored at –70°C until processed.
Samples were batch processed. Vials of frozen sample were thawed, and 20 µl of the fecal broth was subjected to 10-fold dilutions in phosphate-buffered saline. Fifty microliters of the 10–2, 10–4, and 10–6 dilutions was inoculated onto agar medium plates. Selective media for the growth of methicillin-resistant Staphylococcus aureus and vancomycin-resistant enterococci were included in the laboratory work. The agar plates were MacConkey agar (Oxoid Ltd., Basingstoke, United Kingdom), mannitol salt agar with oxacillin (Oxoid Ltd.), and Slanetz & Bartley agar with vancomycin (Oxoid Ltd.). Inoculated plates were incubated at 37°C for 24 to 36 h in air. Gram-negative bacilli that grew on MacConkey agar in air and were resistant to two or more antibiotics were identified. Representatives of different colony types were picked for identification and antibiotic susceptibility testing. Isolates were identified using standard laboratory methods, including the API 20E system (Biomerieux, Basingstoke, United Kingdom).
Antibiotic susceptibility testing, including extended-spectrum β-lactam (ESBL) testing, was carried out using British Society for Antimicrobial Chemotherapy methods and interpretive criteria (www.BSAC.org.uk). The antibiotics tested were cefuroxime, ceftazidime, cefpodoxime, amoxicillin, gentamicin, piperacillin-tazobactam, augmentin, tetracycline, trimethoprim, amikacin, tobramycin, imipenem, ertapenem, tigecycline, colistin, and chloramphenicol. The presence of fully antibiotic-sensitive gram-negative bacilli was recorded, but these strains were not speciated. Cephalosporin resistance was defined as resistance to any of the cephalosporins tested; carbapenem resistance was defined as resistance to ertapenem or imipenem; and aminoglycoside resistance was defined as resistance to amikacin, gentamicin, or tobramycin.
Typing of Escherichia coli and Klebsiella spp. DNA was extracted from discrete colonies growing on blood agar plates using the QIAamp mini kit (Qiagen Ltd., Crawley, United Kingdom) according to the manufacturer's instructions. The DNA concentration of the extract was quantified by spectroscopy, and the concentration was adjusted to 2.5 ng µl–1.
Randomly amplified polymorphic DNA (RAPD) patterns were generated with 5 ng of DNA from each isolate using ready-to-go RAPD analysis beads (Amersham Biosciences) according to the manufacturer's instructions. Primer 2 (5'-GTTTCGCTCC) was used for E. coli, and primer 3 (5'-GTAGACCCGT) was used for Klebsiella spp. Products were analyzed on 2% agarose gels stained with ethidium bromide. Strains were described as indistinguishable if they were indistinguishable by both RAPD and antibiotic susceptibility patterns.
Statistical methods. Multiresistance was defined as resistance to three or more classes of antibiotic. The classes were cephalosporins, amoxicillin, augmentin, tetracycline, trimethoprim, piperacillin-tazobactam, carbapenem, aminoglycoside, colistin, chloramphenicol, and ciprofloxacin.
Logistic regression was used for the analyses using STATA 9; unless otherwise stated, results are from univariate analyses. For the separate classes of antibiotic resistance, the rates of acquisition and loss of a resistant strain between discharge and 6 months after discharge were assumed to be constant throughout the period for each class. The appendix gives further details of the methods used. The two groups, those resistant at discharge and those not, enabled separate differential equations to be given, which, when integrated, equated to the numbers of infants who had resistance at 6 months but did not have resistance at discharge and those who had no resistance at 6 months but did have resistance at discharge. These two equations could be solved (see the appendix) and rates of loss and acquisition estimated. Using these numbers, estimates were made for the proportion of children who had lost or gained resistant strains. The proportions with persisting resistance and their confidence intervals (CI) were calculated with the following formula: 1 – the proportion losing a strain. The CI for these values were calculated by bootstrapping the rates of acquisition and loss and using the confidence limits of the rates to estimate the CI of the proportions at 6 months. Because the formula for estimating acquisition and loss is undefined when one of the rates is very high or very low, the bootstrap method for CI excluded a significant proportion of samples in some cases; these are highlighted.
The proportions of infants actually gaining (or losing) resistance to a class of antibiotic, over the 6 months must be lower than or equal to the proportion of infants who gained (or lost) resistance in this period. This is because some gainers could have then lost resistance and some losers could have gained resistance. If the lower binomial confidence limit from the actual proportions losing or gaining is larger than the bootstrap result or if the bootstrap result is unreliable, then the lower binomial has been substituted. This does not affect the lower CI of the proportion whose resistance persists. Exact CI for the binomial distributions have been used.
|
RESULTS |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXREFERENCES |
|---|
|
Samples were received from 74 (60%) infants at 6 months after discharge, and 47 (38%) were received from infants at 12 months after discharge. Of the 124 infants recruited, 57 infants were discharged before 4 weeks, so that the 2-week sample also was the discharge sample. The proportion carrying strains resistant to each antibiotic at 2 weeks was 79% for tetracycline, 78% for amoxicillin, 31% for cephalosporins, 20% for trimethoprim, 11% for piperacillin-tazobactam, 9% for chloramphenicol, and 4% for aminoglycosides. The proportions of infants colonized with Enterobacteriaceae strains resistant to each antibiotic class at 2 weeks, at discharge, and at 6 and 12 months after discharge are shown in Fig. 1. One strain resistant to tigecycline was detected in the 2-week samples, and no strain was resistant to carbapenems. The majority of infants carrying cephalosporin-resistant strains were carrying cephalosporin-resistant Enterobacter spp. (24 of 38 [63.2%] infants at discharge). Three infants carried ESBL-producing strains at discharge, and four infants carried ESBL-producing strains at 2 weeks.
|
No vancomycin-resistant enterococci were isolated in the discharge samples. A methicillin-resistant S. aureus strain was isolated from samples from one infant at 2 weeks postnatal age and at discharge.
Questionnaires were received for 29 (62%) of the infants from whom stool samples were received at 1 year after discharge. Nine of those 29 infants had received antibiotics, and 6 had been hospitalized in the year following discharge.
There were no statistically significant univariate associations of predetermined risk factors with the carriage of MRE at 2 weeks (Table 2), nor was there any suggestion that low-gestational-age infants were more at risk of colonization with MRE in the first 2 weeks. However, being born at <26 weeks of gestational age was significantly associated with more colonization by MRE at discharge, and in those babies born after 26 weeks the use of a cephalosporin was associated with less MRE colonization at discharge.
|
An investigation of the number of classes of resistance at 2 weeks and at discharge showed that for those colonized with a strain resistant to any antibiotic at 2 weeks, birth at <26 weeks of gestational age was associated with more classes of resistance at discharge (2.4 more classes; 95% CI, 0.7 to 4.0; P = 0.006); however, when adjusted for gestational age, the length of stay was not associated with more classes of resistance at discharge. However, for those without a resistant strain at 2 weeks, the length of stay was independently associated with gaining classes of resistance, with approximately one extra class of resistance per 4 weeks of stay (95% CI, 0.21 to 1.9; P = 0.017).
Persistence of resistant gram-negative bacilli following discharge from the NICU. There was a decline in the proportion of infants carrying strains resistant to cephalosporins, piperacillin-tazobactam, or aminoglycosides following discharge, but not of those carrying strains resistant to amoxicillin, trimethoprim, colistin, tetracycline, or chloramphenicol.
In order to demonstrate the persistence of specific strains, E. coli and Klebsiella spp. were typed by RAPD and antibiotic susceptibility patterns. E. coli was isolated from 34 infants (27.4%) at discharge, from 45 (60.8%) infants at 6 months, and from 27 (57.4%) at 1 year. Strains of E. coli from the 12 infants who were colonized both at discharge and at 6 months were typed. Pairs of E. coli isolates from 9 (75%) of these 12 infants gave indistinguishable RAPD and antibiotic susceptibility patterns, showing that 9 (26.5%; 95% CI, 12.9 to 44%) of 34 infants with E. coli at discharge carried an indistinguishable strain 6 months later. Seven of the 12 infants lost a strain of E. coli, and four gained a new strain. E. coli was isolated from 27 (57.4%) infants at 12 months. Nineteen pairs of isolates from infants colonized at both 6 and 12 months were typed. Eight (42%) of the 19 pairs of isolates were indistinguishable by RAPD and susceptibility patterns, showing that 8 (17.8%; 95% CI, 8 to 32%) of 45 infants carried an indistinguishable strain at 6 and 12 months. Four (11.8%; 95% CI, 3.3 to 27%) of 34 infants who carried E. coli at discharge carried an indistinguishable E. coli isolate at 1 year.
Klebsiella spp. were isolated from 56 (45.2%) infants at discharge, 27 (36.5%) infants at 6 months, and 17 infants (36.2%) at 12 months. Discharge and 6-month pairs of isolates were typed from 16 infants. Of these pairs, four (25%) were indistinguishable by RAPD and antibiotic susceptibility patterns, showing that four (7.1%; 95% CI, 2.0 to 17%) carried indistinguishable strains at discharge and at 6 months. Twelve infants lost a strain and 12 gained a strain between discharge and 6 months.
The proportions of infants carrying an indistinguishable strain of E. coli at discharge and 6 months later was significantly higher than those for Klebsiella spp. (26.5% for E. coli versus 7.1% for Klebsiella spp.; chi-square test, 6.2; P < 0.025).
Estimates of persistence of resistance. Estimates of the extent of persistence were modeled (Table 3) and ranged from 0 to 57% depending on the antibiotic. The estimated levels of persistence were more than 20% for the majority of the antibiotic classes. The estimated persistence rates for augmentin, trimethoprim, tetracycline, chloramphenicol, and piperacillin-tazobactam were high (>33%) and excluded the persistence of less than 20% for trimethoprim, chloramphenicol, and piperacillin-tazobactam.
|
|
DISCUSSION |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXREFERENCES |
|---|
There is an increasing trend toward the use of intrapartum antibiotics for the prevention of GBS disease (35) and to improve the pregnancy outcome in some groups, such as those with preterm rupture of membranes. There are conflicting reports as to whether this has resulted in an increase in the absolute numbers of episodes of early-onset sepsis caused by gram-negative bacteria (25, 33, 37), but there is evidence to suggest that antimicrobial resistance is increasing in those gram-negative isolates associated with both early- and late-onset neonatal sepsis (17, 35, 37). In a report from Washington, DC, 17% of infants admitted to a NICU became colonized with antibiotic-resistant Enterobacteriaceae strains, and 14% of colonized infants developed infection (36). A temporal relationship of colonization preceding bloodstream infection with Enterobacteriaceae strains also has been described (6). In this study, the colonization of infants early in life with MRE was frequent, but despite this there were few episodes of bacteremia associated with MRE (during the inpatient stay).
In a recent study from the United States, 10% of infants presenting to a children's hospital with fever between 57 and 180 days of age were diagnosed with serious bacterial sepsis, of which the great majority were urinary tract infections caused by gram-negative bacilli (most frequently E. coli) (16). In this study, we did not collect detailed information on postdischarge events, although 6 (21%) of 29 infants were reported by caretakers to have been admitted to the hospital in the 12 months following discharge. Antibiotic policies for the treatment of late-onset sepsis in neonates generally recommend an aminoglycoside or a cephalosporin (see, for example, reference 26). These also are mainstays for the treatment of suspected sepsis during the first year of life. In this study, the proportion of infants carrying strains resistant to cephalosporins or aminoglycosides declined after discharge, and cephalosporin- and aminoglycoside-resistant strains were among the least likely to persist. The small proportions carrying ESBL-producing strains may have been an important factor in determining the relatively low level of persistence of cephalosporin resistance in the study population.
In this study, strains resistant to two or more antibiotics were identified using the API system. E. coli and Klebsiella spp. were typed by RAPD. Indistinguishable strains of E. coli were isolated from 26.5% of infants at discharge and at 6 months. High levels of persistence of E. coli in infants have been reported from another northern European country (Sweden) (27), with 53% of infants carrying only one strain of E. coli in the first 6 months of life. The same group (19, 28) reported the persistence of tetracycline-resistant E. coli in Swedish infants, and they also reported that tetracycline-resistant strains were more likely to carry colonization factors.
In this study, estimates of the persistence of Enterobacteriaceae strains were highest for tetracycline- and chloramphenicol-resistant strains. Amoxicillin and related antibiotics frequently are used in infants, but infants are rarely systematically exposed to chloramphenicol or tetracyclines. Chloramphenicol-resistant E. coli strains are widespread in farm animals (3), and this may reflect the association of chloramphenicol resistance with colonization factors (41) and/or linkages with other resistance determinants (39). Compensatory mechanisms that preclude a reversion to antibiotic sensitivity (2, 34) also may be important determinants of persistent resistance to specific antibiotics in populations for which those specific antibiotics are infrequently used.
Many previous studies of antibiotic-resistant gram-negative bacilli have taken place during outbreaks involving single bacterial strains or species in single NICUs and have used a case control design. For example, Linkin et al. reported (23) that, during an outbreak, the risk factors for colonization or infection were low gestational age and exposure to expanded-spectrum cephalosporins. In a study of children during a period of increased isolation of resistant gram-negative bacteria, the administration of an extended-spectrum cephalosporin was reported as a risk factor for infection with a resistant gram-negative bacilli (44). Risk factor analyses for single strains or species may not be generally applicable to antibiotic-resistant Enterobacteriaceae strains, and risk factors for infection may be different from risk factors for colonization. Episodes of colonization and infection frequently are accumulated in analyses. Case control studies may not accurately estimate the extent to which antibiotics are risk factors for the acquisition of resistant strains (14, 30). In particular, case control studies may overemphasize antibiotic exposure as a risk factor for the carriage of resistant strains. The study reported here was a cohort study carried out during a non-outbreak period (as attested to by the low frequency of isolation of MRE from blood cultures during the time period of this study), and we did not find that antibiotic exposure was a risk factor for colonization with resistant Enterobacteriaceae strains in individual infants (apart from piperacillin-tazobactam) when the results from the two units were aggregated. When looked at on the level of individual units, exposure to piperacillin-tazobactam was not a risk factor for a piperacillin-tazobactam-resistant strain. These findings do not preclude the possibility that exposure to specific antibiotics is a risk for colonization with strains resistant to those antibiotics, because high levels of cross-infection may have obscured the effects of antibiotic exposure.
We did find that low gestational age (<26 weeks) was a risk factor for the carriage of antibiotic-resistant Enterobacteriaceae strains at discharge, although it was not a risk factor for acquisition by 2 weeks. Previous studies have reported early colonization and high levels of cross-colonization with antibiotic-resistant gram-negative rods in infants nursed in intensive care units (1, 40). The risk factor findings of this study could be explained if colonization with resistant Enterobacteriaceae strains was a function of the intensity of caretaker-infant interactions and the length of time spent on the unit, with similar levels of intensity of care for infants of various gestational ages in the first 2 weeks following admission.
MRE were found at high frequency in the infants during their stay in the NICUs. Gestational age and antibiotic exposure were not risk factors for the colonization of individual infants with MRE at 2 weeks of age. Antibiotic-resistant strains persisted in a portion of infants. The clinical and public health implications of the carriage and the persistence of carriage of MRE in this population of infants remain to be investigated.
|
APPENDIX |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXREFERENCES |
|---|
(a) Scenario A. In the first scenario, those infants without resistance at the start of the follow-up period give numbers with resistance at the end of the period. Let N(t) be the proportion of infants without resistance at time t. A is the proportion of infants who gained resistance by time 1.
(b) Scenario B. In the second scenario, those infants with resistance at the start of the follow-up period give numbers without resistance at the end of the period. Let R(t) be the proportion with resistance (R) at time t. L is the proportion of infants who lost resistance by time 1.
Equations. The following differential (d) equation is for scenario A: dN = dt[–pN + m(1 – N)] (the proportion of infants acquiring resistance reduces the size of the proportion of infants without resistance, and the proportion losing resistance increases the number of infants without resistance); therefore, dN/dt = m – (p + m)N.
Similarly, for scenario B, dR = dt[–mR + p(1 – R)] (the number of infants losing resistance reduces the number of infants with resistance, and the number gaining resistance increases the number with resistance); therefore, dR/dt = p – (p + m)R.
Applying the integrating factor (IF) method to scenario A yields dN/dt = m – (p + m)N, which can be rearranged to dN/dt + (p + m)N = m. IF = exp[
(p + m)dt] = e(p + m)t (where exp refers to the exponential function). c is a constant.
 |
Applying the IF method to scenario B yields dR/dt = p – (p + m)R, which can be rearranged to dR + (p + m)R = p. For the following equation, IF = exp[(p + m)dt = e(p + m)t] and 1/IF = e–(p + m)t.
 |
When t = 1, R = 1 – L, and N = 1 – A,
 | (1) |
 | (2) |
 |
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Published ahead of print on 26 November 2007.
|
REFERENCES |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Antimicrob. Agents Chemother. | Clin. Microbiol. Rev. |
|---|---|
| Clin. Vaccine Immunol. | ALL ASM JOURNALS |
|---|
