- Email: [email protected]
An antibiotic policy to prevent emergence of resistant bacilli
ARTICLES
An antibiotic policy to prevent emergence of resistant bacilli P de Man, B A N Verhoeven, H A Verbrugh, M C Vos, J N van den Anker
Summary Background Fear of infection in neonatal intensive care units (NICUs) often leads to early use of empiric broad-spectrum antibiotics, a strategy that selects for resistant bacteria. We investigated whether the emergence of resistant strains could be halted by modifying the empiric antibiotic regimens to remove the selective pressure that favours resistant bacteria. Methods Two identical NICUs were assigned to different empiric antibiotic regimens. On unit A, penicillin G and tobramycin were used for early-onset septicaemia, flucloxacillin and tobramycin were used for late-onset septicaemia, and no broad-spectrum -lactam antibiotics, such as amoxicillin and cefotaxime were used. In unit B, intravenous amoxicillin with cefotaxime was the empiric therapy. After 6 months of the study the units exchanged regimens. Rectal and respiratory cultures were taken on a weekly basis. Findings There were 436 admissions, divided equally between the two regimens (218 in each). Three neonates treated with the penicillin-tobramycin regimen became colonised with bacilli resistant to the empirical therapy used versus 41 neonates on the amoxicillin-cefotaxime regimen (p<0·001). The relative risk for colonisation with strains resistant to the empirical therapy per 1000 patient days at risk was 18 times higher for the amoxicillin-cefotaxime regimen compared with the penicillin-tobramycin regimen (95% CI 5·6–58·0). Enterobacter cloacae was the predominant bacillus in neonates on the amoxicillincefotaxime regimen, whereas Escherichia coli predominated in neonates on the penicillin-tobramycin regimen. These colonisation patterns were also seen when the units exchanged regimens. Interpretation Policies regarding the empiric use of antibiotics do matter in the control of antimicrobial resistance. A regimen avoiding amoxicillin and cefotaxime restricts the resistance problem. Lancet 2000, 355: 973–78
Introduction More preterm infants in need of aggressive invasive therapies are being admitted to neonatal intensive care units (NICUs) than ever before. Infection, and fear of infection in this setting often leads to early use of empiric broad-spectrum antibiotics, a strategy that may well select for resistant bacteria.1 For babies in NICUs, selection of resistant bacteria is facilitated by the fact that new patients enter NICUs directly after birth, often after delivery by caesarean section, with little exposure to bacteria. These babies are therefore at great risk to become colonised by ward strains, which are often resistant. Empiric antibiotic regimens for suspected septicaemia in neonates need to be able to treat group B streptococci, Listeria spp, and Escherichia coli. The regimens commonly include a broad-spectrum penicillin, such as ampicillin or amoxicillin, combined with an aminoglycoside or a thirdgeneration cephalosporin. Amoxicillin exerts a selective pressure towards -lactamase-producing bacteria such as Klebsiella spp. Amoxicillin-containing antibiotic regimens in NICUs have been thwarted by outbreaks of resistant klebsiella.2–7 Many reports have shown that thirdgeneration cephalosporins select strains of Enterobacter spp and Serratia spp, which contain chromosomal lactamase genes.7–11 Use of third-generation cephalosporins has also been linked to the emergence of bacteria that produce plasmid-coded, extended-spectrum -lactamases in NICUs and other wards.12 During the past few years our NICUs have been repeatedly confronted with the emergence of resistant gram-negative bacilli associated with the empiric use of amoxicillin-cefotaxime combination therapy for suspected neonatal sepsis. Therefore, we felt the need to modify our empiric antibiotic regimen, and decided to do so in a wellcontrolled manner. We, thus, compared our amoxicillincefotaxime regimen with an alternative penicillintobramycin regimen that is also appropriate therapy for this indication, but does not involve antibiotics that select for resistant gram-negative bacilli. This change in antibiotic policy was done as a prospective cross-over intervention trial in two identical units for neonatal intensive care.
Methods Participants and procedures
Department of Medical Microbiology and Infection Control, Erasmus University Medical Centre (P de Man MD, B A N Verhoeven MD, Prof H A Verbrugh MD, M C Vos MD); Department of Medical Microbiology and Infection Control, Sint Franciscus Gasthuis (P de Man); and Department of Paediatrics, Sophia’s Children’s Hospital, Erasmus University Medical Centre, Rotterdam, Netherlands (Prof J N van den Anker MD) Correspondence to: Dr P de Man, Department of Medical Microbiology and Infection Control, Sint Franciscus Gasthuis, PO Box 10900, 3004 BA, Rotterdam, Netherlands (e-mail: [email protected])
THE LANCET • Vol 355 • March 18, 2000
Our study was done in the division of neonatology, Sophia Children’s Hospital, which has three units. Two of these provide intensive care with mechanical ventilation. One of these two wards (unit A) has eight incubators whereas the other (unit B) has nine; otherwise the units are identical. Neonates no longer in need of this type of care are moved to a third unit without ventilator facilities. The two intensive units are adjacent to each other with a communicating door. Although unit A and B have separate teams of health-care workers, staff do sometimes use this door to work on the other unit. The effects of a change in antibiotic policy were studied as a prospective intervention, in which the two units were using different empiric antibiotic regimens. During the first 6 months of the study, unit A used intravenous amoxicillin combined with cefotaxime (amoxicillin-cefotaxime regimen) and unit B used
973
ARTICLES Penicillintobramycin regimen Number of admissions Measurements on admission Birthweight (g) Gestational age (weeks) Proportion delivered by caesarean section (%) Apgar at 1 min* (n=389) Apgar at 5 min† (n=390) Outcome measurements Weight at transfer (g) Time in unit (days) Number who died Proportion of days on ventilator (%) Proportion of days with central venous line (%)
218
Amoxicillincefotaxime regimen 218
1935 (1009) 33·3 (4·5) 42·4%
2101 (1104) 33·3 (4·5) 41·6%
5·6 (2·6) 7·7 (1·9)
6·0 (2·5) 7·9 (1·7)
2033 (921) 11·3 (16·5) 26 (11·9%) 40·4% 34·7%
p
2120 (980) 12·9 (20·5) 16 (7·3%) 40·0% 39·0%
0·2 0·6 0·9 0·9 0·23 0·4 0·001 0·14 0·7 0·005
Data are mean (SD) unless indicated. *Only 389 neonates were scored. †Only 390 neonates were scored.
Table 1: Characteristics of newborn babies penicillin G and tobramycin (penicillin-tobramycin regimen). After 6 months the units switched antibiotic regimens for a further 6 months (cross-over phase) to balance the effect of possible unit-related confounding variables. The chronic shortage of NICU beds in relation to the number of neonates made it impossible to randomly assign neonates to one of the treatment groups on admission. In practice, new neonates were admitted to the first available location, a vacancy often created by a transfer of another neonate to the non-ventilator unit. On the rare occasion where both ventilator units had capacity for new admission, the neonate was admitted to the unit with the lowest occupancy at that time. All new patients admitted to the two units were included in the analysis. Patients leaving the unit on the day of admission were not included in the analysis. The penicillin-tobramycin regimen involved these two antibiotics as empiric therapy for episodes of suspected septicaemia in the first 48 h of life (early onset). Flucloxacillin, which is also a narrow-spectrum penicillin, combined with tobramycin, was used for late-onset septicaemia (beyond the first 48 h of life). To avoid any use of amoxicillin and cephalosporins during the penicillin-tobramycin regimen, meropenem was used for culture-proven gram-negative bacteraemia. Meropenem was also prescribed for patients with suspected meningitis since there are no data for newborn babies on aminoglycoside penetration into the central nervous system. The regimen with broadspectrum penicillins and cephalosporins, the amoxicillincefotaxime regimen, used these two antibiotics for early-onset septicaemia. For late-onset septicaemia cefotaxime was combined with flucloxacillin. In month 4 of the study, after two cases of cefotaxime-resistant enterobacter bacteraemia had occurred, flucloxacillin was replaced by tobramycin for neonates with suspected late-onset septicaemia. Both regimens used vancomycin for persistent septicaemia in the presence of central venous lines. Neonates being treated with antibiotics at the time of crossover of the antibiotic regimens continued on their own regimens, but new courses of antibiotics were given in accordance with the prevailing regimen. In the analysis of the neonate-related measurements, such as Apgar scores and birthweight, the neonates were counted according to the regimen on admission. Antibiotic use was scored in relation to the regimen of the unit at that time. Both antibiotic regimens used in our study are routinely used in the Netherlands. We monitored a controlled change in antibiotic policy that did not involve investigational drugs or additional sampling. The study was discussed with the ethics committee. The committee considered this study to be a qualitycontrol investigation of the hospital, and not experimentation with human beings that would require formal ethics review and informed consent. However, parents and guardians were informed of all antibiotic treatments.
974
Colonisation strain
Penicillintobramycin regimen
Amoxicillincefotaxime regimen
Gram-negative bacillus resistant to cefotaxime Gram-negative bacillus resistant to tobramycin Gram-negative bacillus resistant to cefotaxime or tobramycin Cefotaxime-resistant Enterobacter spp Gram-negative bacillus resistant to empiric therapy of unit*
6·8 (16/2339)
21·4 (41/1914)
1·2 (3/2519)
0 (0/2706)
Relative risk (95% CI) 3·14 (1·76–5·56) ··
8·9 (19/2128)
21·4 (41/1914)
2·42 (1·41–4·15)
6·8 (15/2197)
20·3 (39/1917)
2·98 (1·64–5·38)
1·2 (3/2519)
21·4 (41/1914)
17·98 (5·57–58·01)
Data are (colonising events/patient days at risk) ⫻1000. *Tobramycin resistant in unit using penicillin-tobramycin regimen and cefotaxime resistant in unit using amoxicillin-cefotaxime regimen.
Table 2: Risk of colonisation with resistant gram-negative bacilli and antibiotic regimen Data on the neonates were collected on case record forms. For every neonate one form was used to record measurements such as Apgar scores and birthweight. Another case record form was used for each day of the admission with data such as antibiotic use, presence of central venous line, and type of ventilation for that particular day. Bacterial screening included cultures of respiratory aspirates and rectal swabs at admission and once a week thereafter. Samples were inoculated onto MacConkey and blood-agar plates. Plates were inspected after 18 h and 42 h of incubation at 37ºC. Each morphologically different colony type of potential pathogens, including all gram-negative bacilli, was selected for further identification. The number of different gram-negative bacteria is limited in these types and patients, usually there were only one or two different colony types and never more than four. When biochemistry and antibiograms of these colonies were identical, they were subsequently reported as one strain. Identification was carried out, and antibiograms were done with the MicroScan system (Dade International, West Sacramento, USA). Additional samples such as respiratory aspirates, blood, and cerebrospinal fluids were taken only when clinically indicated. Hygienic precautions included strict regulations for clothing, and hand disinfection with 70% ethanol was obligatory in between contact with each neonate.
Statistical analysis The data were analysed with the aid of the SAS system (version 6.12). Comparisons of means were done with the Student’s t test, and comparisons of proportion with Fisher’s exact test. The CIs for the relative risks were calculated with the Katz approximation. Colonisation with resistant bacteria was the primary outcome measurement of the effect of the change in antibiotic policy. The other comparisons were done to assess possible wanted and unwanted secondary effects of the changes in antibiotic regimen. Readers should be aware of the risk of significance by chance in these many comparisons.
Results From December, 1996, to December, 1997, 480 neonates were admitted to the two NICUs. 44 neonates left the unit on the day of admission and, therefore, were not included in this analysis. By chance, the remaining 436 admissions divided equally over the two regimens (218 admitted to each regimen). The neonates’ measurements on admission did not differ significantly between the two regimens (table 1). However, there were substantial differences between the two regimens with regard to the rate of colonisation with gram-negative bacilli. During the penicillin-tobramycin regimen E coli predominated and made up 53% of all gram-negative isolates. For the amoxicillin-cefotaxime regimen, 77% of all gram-negative isolates were Enterobacter spp. The species distribution of
THE LANCET • Vol 355 • March 18, 2000
ARTICLES
NICU A Isolates of bacilli resistant to empiric antibiotic regimen of unit Isolates of bacilli sensitive to empiric antibiotic regimen of unit but resistant to antibiotic regimen of other unit
1
4
8
12
16
20
24
28
32
Amoxicillin-cefotax ime regimen
36
40
44
48
52
48
52
Penicillin-tobramycin regimen
NICU B
1
4
8
12
16
20
24
28
Penicillin-tobramycin regimen
32
36
40
44
Amox icillin-cefotax ime regimen
Week number Isolation distribution of resistant bacilli from patients according to the antibiotic regimen
the remaining gram-negative isolates was not associated with the empiric antibiotic regimen. The risk of colonisation with resistant gram-negative bacilli expressed as colonising events per 1000 patient days at risk was calculated for five separate patterns of resistance. Emergence of resistance was higher during the amoxicillin-cefotaxime regimen (table 2). There were more episodes of colonisation with either tobramycinresistant or cefotaxime-resistant strains and there were more cefotaxime-resistant enterobacter among neonates on the amoxicillin-cefotaxime regimen. The most striking, and probably most relevant difference was found in the analysis of colonising events with strains resistant to the agents of the empirical antibiotic regimen in use on the unit. Colonising events with such strains occurred 18 times more frequently in the units when the amoxicillincefotaxime regimen was used than when the penicillintobramycin regimen was used (table 2). The development of resistance over time in the two units is shown in the figure. The colonisation with bacilli resistant to the empiric therapy is a reflection of antibiotic regimens, and closely follows the cross-over of antibiotic policy between the two NICUs. Colonisation with pathogens other than gram-negative bacilli, such as Candida spp, enterococci, Haemophilus spp, Listeria monocytogenes, coagulase-negative staphylococci, Staphylococcal aureus, and group-A and group-B streptococci, did not differ between the two
THE LANCET • Vol 355 • March 18, 2000
regimens (data not shown). This finding indicates that the problem with resistant gram-negative bacilli was not replaced by problems with other pathogens in the penicillin-tobramycin regimen. Despite the fact that 285 blood cultures were done within the first 48 h of life, there was a remarkable absence of culture-proven group-B streptococci bacteraemia. All cases of E coli bacteraemia occurred in neonates after day 6 of life. One neonate was colonised with L monocytogenes, which was not cultured from deep sites. Thus, there were no cases of early-onset bacterial neonatal sepsis or meningitis detected. However, there were nosocomial infections as complications of the Organism
Penicillintobramycin regimen
Amoxicillincefotaxime regimen
p
Candida spp Coagulase-negative staphylococci† Micrococcus luteus Staphylococcus aureus Enterococci Streptococci Bacillus cereus Enterobacter cloacae Klebsiella oxytoca Escherichia coli Total
2 15 0 0 0 2 0 0 1 7 27
1 23 2 1 2 1 1 5 0 1 37
. .* 0·23 ··* ··* ··* ··* ··* 0·06 ··* 0·07 0·22
Data are number of episodes of bacteraemia. *p values were not estimated for infections that occurred less than five times. †All but one of the coagulase-negative staphylococci were methicillin resistant in each regimen.
Table 3: Episodes of bacteraemia by causative organism
975
ARTICLES
Antibiotic
Penicillin-tobramycin regimen
Amoxicillin-cefotaxime regimen
Indication
% of Indication days* 28
Amoxicillin
Early-onset septicaemia Early and lateonset septicaemia None†
Cefotaxime
None†
Flucloxacillin
Late-onset septicaemia Culture proven gram-negative bacteraemia Late-onset septicaemia with central venous line
Penicillin Tobramycin
Meropenem
Vancomycin
34·7
None†
Late-onset septicaemia 1·2† Early-onset septicaemia 1·5† Early and lateonset septicaemia 9·8 Late-onset septicaemia 7·6 Cefotaxime-resistant gram-negative bacteraemia 3·2 Late-onset septicaemia with central venous line
p % of days* 0·8†
··
4
··
25·6
··
31·7
··
8·3
0·67
6·4
0·64
5·7
<0·001
*Number of days treated with antibiotic divided by total number of days multiplied by 100. †Used during the cross-over of regimens.
Table 4: Antibiotic use
intensive-care treatment. Differences in the rate of bacteraemia did not reach the 0·05 level of significance (table 3). However, enterobacter bacteraemia occurred five times in neonates on the amoxicillin-cefotaxime regimen but not at all in babies on the penicillintobramycin regimen (p=0·06). Conversely, E coli bacteraemia was seen more often with the penicillintobramycin regimen than with the amoxicillin-cefotaxime regimen (7 vs 1, p=0·07). There were no cases of tobramycin-resistant gramnegative bacteraemia among neonates on the penicillintobramycin regimen, indicating that the neonates on this regimen were always protected by these empiric antibiotics in case septicaemia occurred. There were four cases of cefotaxime-resistant gram-negative bacteraemia in neonates on the amoxicillin-cefotaxime regimen. After the first two cases occurred, the antibiotics for late-onset septicaemia used in this regimen were changed from flucloxacillin with cefotaxime to tobramycin with cefotaxime. Antibiotic use was a reflection of the empiric antibiotic regimens (table 4). The low-level use of penicillin during the amoxicillin-cefotaxime regimen and use of amoxicillin and cefotaxime during the penicillin-tobramycin regimen was a result of neonates continuing their course of antibiotics after the unit cross-over. There was one (short lived) violation of the protocol when a neonate in the penicillin-tobramycin regimen was given amoxicillin for an E coli infection. Remarkably, the use of meropenem, which was empiric therapy for meningitis and standard treatment for all cases of culture-proven gram-negative bacteraemia for neonates on the penicillin-tobramycin regimen, did not differ from the amoxicillin-cefotaxime regimen where it was only intended for use in cefotaximeresistant bacteraemia. Two outcome measurements differed significantly between the two regimens—the duration of stay in the unit was on average 1·5 days shorter for neonates treated with the penicillin-tobramycin regimen, and the use of central venous lines was lower for the penicillin-tobramycin regimen (table 1). This observation coincides with a higher vancomycin use in neonates on the amoxicillin-cefotaxime regimens. The other measurements did not differ significantly between the two regimens. The causes of death are listed in table 5. There was only one infection-related death, which occurred in a girl on chemotherapy for neuroblastoma. She had a central
976
Cause of death
Penicillintobramycin regimen
Amoxicillincefotaxime regimen
Asphyxia Encephalopathy Microencephalopathy Intracerebral bleeding Convulsions Twin to twin transfusion Neuroblastoma with Enterobacter sepsis Severe combined immunodeficiency syndrome Pulmonary hypoplasia Pulmonary bleeding Cystic lung malformation Pulmonary hypertension Meconium aspiration syndrome Hypoplastic left ventricle Tetralogy of Fallot Pneumopericardium Triploidy Pyruvate dehydrogenase deficiency Unknown Total
5 2 1 4 1 2 0 0 6 0 0 1 1 1 1 0 1 0 0 26*
4 0 0 0 0 0 1 1 3 1 1 0 0 1 1 1 0 1 1 16†
*There were no deaths attributed to renal failure. †There were no significant differences.
Table 5: Causes of death
venous line and developed sepsis with cefotaximeresistant enterobacter. She remained line dependent because of septic shock. Although the enterobacter was sensitive towards meropenem and tobramycin, treatment with these drugs could not reverse the bacteraemic state in the presence of the line. The other fatal cases occurred in patients with major underlying diseases or abnormalities. Careful re-assessment of the patients’ files did not show a role for infection in those cases. The amoxicillin-cefotaxime regimen was designed to use cefotaxime with flucloxacillin for late-onset septicaemia. However, in the fourth month of the study two neonates that developed this disease entity were treated with meropenem because they were known to be colonised with cefotaxime-resistant enterobacter. They recovered, and indeed their blood cultures were positive with these resistant bacteria. After these incidents, the study protocol was changed to contain tobramycin for all neonates developing late-onset septicaemia, since colonisation cultures do not always precede sepsis and we wanted to guarantee the best treatment for all children. These interventions however, may have concealed the effect of chance in antibiotic policy on mortality, since they may well have caused a decrease in mortality in the amoxicillin-cefotaxime regimen.
Discussion This study shows that policies about the empiric use of antibiotic regimens do matter in the control of antimicrobial resistance in an intensive-care setting. The flora of the body surfaces forms the reservoir for invasive infections. We found a 18-fold higher rate of colonisation with strains resistant to the empiric therapy in patients on the amoxicillin-cefotaxime regimen than in those on penicillin-tobramycin regimen. The newborn babies treated with the penicillin-tobramycin regimen were thus better protected against nosocomial septicaemia with colonising gram-negative strains than those who were routinely treated with amoxicillin and cefotaxime. In this paper we do not describe an analysis of the factors associated with the risk for colonisation with resistant bacteria in these neonates, other than the antibiotics used. We did do such an analysis, which showed that risk for colonisation was related to low
THE LANCET • Vol 355 • March 18, 2000
ARTICLES
gestational age, low birthweight, and the length of stay, factors known from previous studies. Numerous other factors did not contribute. The analysis was done as a case-control study, considerably different from the cohort study presented in this paper, and therefore will be published elsewhere. The number of prospective studies correlating antibiotic regimens and bacterial resistance is limited, and there is no absolute proof of a causative association between antibiotic use and resistance.13 The most relevant information on this subject originates from reports on the management of outbreaks and other studies with historical controls. Price and Sleigh14 halted an outbreak of resistant klebsiella by a complete cessation of all antibiotic prescribing in a neurosurgical intensive-care unit. In a preintervention versus postintervention study in a paediatric intensive-care unit, Toltzis and co-workers15 were able to reduce ceftazidime use by 96% but were confronted with an increase in ceftazidime resistance. However, White and colleagues16 restricted antibiotic use in the management of an outbreak of resistant acinetobacter and achieved a significant decrease not only in resistant acinetobacter but also in the antibiotic resistance of Enterobacteriaceae. In Finland a nationwide reduction of macrolide use was associated with a reduction in macrolide resistance among group-A haemolytic streptococci.17 The antibiotics in the traditional regimens in NICUs have been selected for their activity against early-onset neonatal septicaemia with pathogens acquired in utero or during labour. However, there have been changes in NICUs over the years. The admissions have shifted towards more premature infants often delivered by caesarean section for maternal reasons not related to infection. These premature infants are in need of intensive care over extended periods of time. They are seldom born with infection but are at increasing risk of acquiring infection during their stay in the NICU.18 A remarkable observation in this study was the absence of typical earlyonset neonatal septicaemia. An antibiotic regimen should not only treat the normal pathogens of neonatal septicaemia but it should also provide protection against the nosocomial pathogens that threaten the newborn babies during their stay in the NICU. Antibiotic treatment by the classic antibiotic regimens in NICU, tends to select bacteria with resistance towards the antibiotics used. In the amoxicillin-aminoglycoside regimens there is a selective pressure towards -lactamase producers that can degrade the amoxicillin. Such strains including klebsiellae have a clear advantage over normally sensitive bacteria such as E coli, and are, therefore, likely to become resident ward strains. These ward strains are then repeatedly exposed to the aminoglycosides and in turn may become aminoglycoside-resistant strains with outbreak potential.2–7 Cephalosporins evade the problems of resistance because they have less effect on the normal intestinal flora.19 This theory is supported by the observation that neonates treated with amoxicillin had more resistant gram-negative bacilli than neonates who had been on cefuroxime treatment.20–23 In the cefotaxime-containing regimens used in NICUs including ours, cefotaxime was combined with amoxicillin to cover Listeria spp. In these circumstances the amoxicillin will eliminate the normal intestinal flora. Gram-negative bacteria such as Enterobacter spp and Serratia spp that can degrade
THE LANCET • Vol 355 • March 18, 2000
cefotaxime and amoxicillin will be selected and may flourish.7,24,25 Resistant enterobacter strains will cause invasive infections in NICU and cannot be treated with cefotaxime.8 The resistance of enterobacter has been seen to reach a plateau at 25–35% resistant strains. This is explained by the fact that most enterobacter infections arise from the patients own endogenous flora. This argument clearly does not hold for NICU, where the ward flora is the reservoir for both colonisation and infection. The penicillin-tobramycin regimen differs from these two regimens by the fact that it does not include a lactam with a selective effect towards resistant gramnegative bacilli. Penicillin G does not penetrate the outer membrane of gram-negative bacilli and, therefore, lacks activity towards these bacteria. Thus, it does not generate a selective pressure towards resistant, -lactamaseproducing bacilli. Meropenem was the only broadspectrum -lactam we used in the penicillin-tobramycin regimen. It was chosen as a treatment for culture positive gram-negative bacteraemia, because it lacks a selective effect towards bacteria such as Enterobacter spp and Serratia spp—largely due to its stability towards most lactamases. However, an overuse of this drug can evoke problems with strains of Stenothrophomonas maltophilia, which produces -lactamase capable of inactivating carbapenems. Use of aminoglycoside in the penicillin-tobramycin regimens was not associated with a significant emergence of resistance during the study (only three colonising events in 2515 patient days). Aminoglycosides, given parenterally, do not penetrate the mucosal surfaces of the gastrointestinal tract and therefore may have a low selective effect. In fact, our study and others26 show that newborn babies on aminoglycoside therapy become colonised with gram-negative bacteria sensitive to these antibiotics. The choice of -lactams that have no selective effects—either alone or in a combination with an aminoglycoside that does not penetrate mucosal membranes—may be the reason for the striking reduction in resistance seen in this study. Penicillin G and tobramycin do cover the pathogens of early-onset neonatal septicaemia, since almost all community acquired E coli are sensitive to tobramycin and group-B streptococci and Listeria spp are sensitive to penicillin G. The empiric treatment of suspected bacterial meningitis remains a problem, since the penetration of aminoglycosides into the central nervous system is poor in adults. The undeveloped blood-brain barrier in preterm infants might allow for more effective penetration. Since we are not aware of studies on this subject, we used meropenem when meningitis was suspected. The high amount of antibiotic use, in combination with the low grade of colonisation of neonates at the time of their admissions, turns the NICU into an environment where antibiotic policy is likely to have a pronounced effect on the resistance problem. These facts contribute to the clear differences found in this study. Although in other wards the effects of antibiotic therapy can be less obvious, the mechanisms behind the selection of resistant nosocomial strains are probably similar if not identical. In fact, Enterobacter spp and other -lactamase producing bacteria are frequently encountered throughout the hospital.27 The approach of this change in policy, avoiding broad-spectrum penicillins and cephalosporins could be used in other wards that have problems with resistant Enterobacteriaceae, or such a change can be seen as a
977
ARTICLES
preventive measure that will be of value in wards that use high quantities of antibiotics. We have been using the penicillin-tobramycin regimen for neonatal sepsis in all our NICUs for 2 years since the end of the study. There have been no further or new problems with resistance or treatment failures in this period. This study provides solid evidence of a reduction of bacterial resistance mediated by the use of antibiotics that exert little selective pressure on the mucosal environment, confirming the benefits of the often preached but less frequently practised use of narrow-spectrum antibiotics. This study might inspire clinicians to put that old knowledge into practice more often.
9
Contributors
14
P de Man did the initial protocol design, coordinated the microbiology, and contributed to the statistical analysis and the writing of the paper. B A N Verhoeven coordinated and collected clinical data and contributed to the statistical analysis and writing of the paper. H A Verbrugh contributed to the protocol design, the analysis, and the writing of the paper. M C Vos contributed to the writing of the paper. J N van den Anker coordinated the clinical aspects of the study, reviewed the charts of all fatalities, designed and analysed the study, and wrote the paper.
10
11
12
13
15
16
Acknowledgments We thank the nursing staff of the department of neonatology, the laboratory technicians, and A Smit of the department of medical microbiology and infectious diseases for their conscientious work during this study; D van Dongen for collection of clinical data; M Vogel for extracting computer information from the hospital system; J G M Koeleman, J W Mouton, C Verduin, and I C Gyssens for discussion and proofreading; and D M MacLaren for his original idea to avoid all selective effects of -lactams by the use of penicillin G instead of amoxicillin in the control of a serratia outbreak in 1994 (NICU of the Free University, Amsterdam, Netherlands).
17
References
20
1
2
3
4
5
6
7
8
Hammerschlag MR, Klein JO, Herschel M, Chen FC, Fermin R. Patterns of use of antibiotics in two newborn nurseries. N Engl J Med 1977; 296: 1268–69. Noy JH, Ayliffe GA, Linton KB. Antibiotic-resistant gram-negative bacilli in the faeces of neonates. J Med Microbiol 1974; 7: 509–20. Mayhall CG, Lamb VA, Bitar CM, et al. Nosocomial klebsiella infection in a neonatal unit: identification of risk factors for gastrointestinal colonization. Infect Control 1980; 1: 239–46. Markowitz SM, Veazey JM Jr, Macrina FL, Mayhall CG, Lamb VA. Sequential outbreaks of infection due to Klebsiella pneumoniae in a neonatal intensive care unit: implication of a conjugative R plasmid. J Infect Dis 1980; 142: 106–12. Kayyali MZ, Nicholson DP, Smith IM. A klebsiella outbreak in a pediatric nursery: emergency action and preventive surveillance. Clin Pediatr (Phila) 1972; 11: 422–26. Arredondo-Garcia JL, Diaz-Ramos R, Solorzano-Santos F, Sosa-Gonzalez IE, Beltran-Zuniga M. Septicaemia due to Klebsiella pneumoniae in newborn infants: nosocomial outbreak in an intensive care unit. Rev Latinoam Microbiol 1992; 34: 11–16. Bryan CS, John JF Jr, Pai MS, Austin TL. Gentamicin vs cefotaxime for therapy of neonatal sepsis: relationship to drug resistance. Am J Dis Child 1985; 139: 1086–89. Modi N, Damjanovic V, Cooke RW. Outbreak of cephalosporin resistant Enterobacter cloacae infection in a neonatal intensive care unit. Arch Dis Child 1987; 62: 148–51.
978
18
19
21
22
23
24
25
26
27
Haertl R, Bandlow G. Molecular typing of Enterobacter cloacae by pulsed-field gel electrophoresis of genomic restriction fragments. J Hosp Infect 1993; 25: 109–16. Spritzer R, Kamp HJ, Dzoljic G, Sauer PJ. Five years of cefotaxime use in a neonatal intensive care unit. Pediatr Infect Dis J 1990; 9: 92–96. Verweij PE, Van Belkum A, Melchers WJ, Voss A, HoogkampKorstanje JA, Meis JF. Interrepeat fingerprinting of third-generation cephalosporin-resistant Enterobacter cloacae isolated during an outbreak in a neonatal intensive care unit. Infect Control Hosp Epidemiol 1995; 16: 25–29. Venezia RA, Scarano FJ, Preston KE, et al. Molecular epidemiology of an SHV-5 extended-spectrum beta-lactamase in enterobacteriaceae isolated from infants in a neonatal intensive care unit. Clin Infect Dis 1995; 21: 915–23. Gould IM. A review of the role of antibiotic policies in the control of antibiotic resistance. J Antimicro Chemoth 1999; 43: 459–65. Price DJ, Sleigh JD. Control of infection due to Klebsiella aerogenes in a neurosurgical unit by withdrawal of all antibiotics. Lancet 1970; ii: 1213–15. Toltzis P, Yamashita T, Vilt L, et al. Antibiotic restriction does not alter endemic colonization with resistant gram-negative rods in a pediatric intensive care unit. Crit Care Med 1998; 26: 1893–99. White AC Jr, Atmar RL, Wilson J, Cate TR, Stager CE, Greenberg SB. Effects of requiring prior authorization for selected antimicrobials: expenditures, susceptibilities, and clinical outcomes. Clin Infect Dis 1997; 25: 230–39. Seppala H, Klaukka T, Vuopio-Varkila J, et al. The effect of changes in the consumption of macrolide antibiotics on erythromycin resistance in group A streptococci in Finland. Finnish Study Group for Antimicrobial Resistance. N Engl J Med 1997; 337: 441–46. Khadilkar V, Tudehope D, Fraser S. A prospective study of nosocomial infection in a neonatal intensive care unit. J Paediatr Child Health 1995; 31: 387–91. Ledingham IM, Alcock SR, Eastaway AT, McDonald JC, McKay IC, Ramsay G. Triple regimen of selective decontamination of the digestive tract, systemic cefotaxime, and microbiological surveillance for prevention of acquired infection in intensive care. Lancet 1988; i: 785–90. Tullus K, Burman LG. Ecological impact of ampicillin and cefuroxime in neonatal units. Lancet 1989; i: 1405–07. Tullus K, Berglund B, Berman LG. Emergence of cross-resistance to beta-lactam antibiotics in fecal Escherichia coli and Klebsiella strains from neonates treated with ampicillin or cefuroxime. Antimicrob Agents Chemother 1990; 34: 361–62. Burman LG, Berglund B, Huovinen P, Tullus K. Effect of ampicillin versus cefuroxime on the emergence of beta-lactam resistance in faecal Enterobacter cloacae isolates from neonates. J Antimicrob Chemother 1993; 31: 111–16. Kalenic S, Francetic I, Polak J, Zele-Starcevic L, Bencic Z. Impact of ampicillin and cefuroxime on bacterial colonization and infection in patients on a neonatal intensive care unit. J Hosp Infect 1993; 23: 35–41. Acolet D, Ahmet Z, Houang E, Hurley R, Kaufmann ME. Enterobacter cloacae in a neonatal intensive care unit: account of an outbreak and its relationship to use of third generation cephalosporins. J Hosp Infect 1994; 28: 273–86. Berkowitz FE, Metchock B.Third generation cephalosporin-resistant gram-negative bacilli in the feces of hospitalized children. Pediatr Infect Dis J 1995; 14: 97–100. Blakey JL, Lubitz L, Barnes GL, Bishop RF, Campbell NT, Gillam GL. Development of gut colonisation in pre-term neonates. J Med Microbiol 1982; 15: 519–29. Sanders WE Jr, Sanders CC. Enterobacter spp: pathogens poised to flourish at the turn of the century. Clin Microbiol Rev 1997; 10: 220–41.
THE LANCET • Vol 355 • March 18, 2000
An antibiotic policy to prevent emergence of resistant bacilli P de Man, B A N Verhoeven, H A Verbrugh, M C Vos, J N van den Anker
Summary Background Fear of infection in neonatal intensive care units (NICUs) often leads to early use of empiric broad-spectrum antibiotics, a strategy that selects for resistant bacteria. We investigated whether the emergence of resistant strains could be halted by modifying the empiric antibiotic regimens to remove the selective pressure that favours resistant bacteria. Methods Two identical NICUs were assigned to different empiric antibiotic regimens. On unit A, penicillin G and tobramycin were used for early-onset septicaemia, flucloxacillin and tobramycin were used for late-onset septicaemia, and no broad-spectrum -lactam antibiotics, such as amoxicillin and cefotaxime were used. In unit B, intravenous amoxicillin with cefotaxime was the empiric therapy. After 6 months of the study the units exchanged regimens. Rectal and respiratory cultures were taken on a weekly basis. Findings There were 436 admissions, divided equally between the two regimens (218 in each). Three neonates treated with the penicillin-tobramycin regimen became colonised with bacilli resistant to the empirical therapy used versus 41 neonates on the amoxicillin-cefotaxime regimen (p<0·001). The relative risk for colonisation with strains resistant to the empirical therapy per 1000 patient days at risk was 18 times higher for the amoxicillin-cefotaxime regimen compared with the penicillin-tobramycin regimen (95% CI 5·6–58·0). Enterobacter cloacae was the predominant bacillus in neonates on the amoxicillincefotaxime regimen, whereas Escherichia coli predominated in neonates on the penicillin-tobramycin regimen. These colonisation patterns were also seen when the units exchanged regimens. Interpretation Policies regarding the empiric use of antibiotics do matter in the control of antimicrobial resistance. A regimen avoiding amoxicillin and cefotaxime restricts the resistance problem. Lancet 2000, 355: 973–78
Introduction More preterm infants in need of aggressive invasive therapies are being admitted to neonatal intensive care units (NICUs) than ever before. Infection, and fear of infection in this setting often leads to early use of empiric broad-spectrum antibiotics, a strategy that may well select for resistant bacteria.1 For babies in NICUs, selection of resistant bacteria is facilitated by the fact that new patients enter NICUs directly after birth, often after delivery by caesarean section, with little exposure to bacteria. These babies are therefore at great risk to become colonised by ward strains, which are often resistant. Empiric antibiotic regimens for suspected septicaemia in neonates need to be able to treat group B streptococci, Listeria spp, and Escherichia coli. The regimens commonly include a broad-spectrum penicillin, such as ampicillin or amoxicillin, combined with an aminoglycoside or a thirdgeneration cephalosporin. Amoxicillin exerts a selective pressure towards -lactamase-producing bacteria such as Klebsiella spp. Amoxicillin-containing antibiotic regimens in NICUs have been thwarted by outbreaks of resistant klebsiella.2–7 Many reports have shown that thirdgeneration cephalosporins select strains of Enterobacter spp and Serratia spp, which contain chromosomal lactamase genes.7–11 Use of third-generation cephalosporins has also been linked to the emergence of bacteria that produce plasmid-coded, extended-spectrum -lactamases in NICUs and other wards.12 During the past few years our NICUs have been repeatedly confronted with the emergence of resistant gram-negative bacilli associated with the empiric use of amoxicillin-cefotaxime combination therapy for suspected neonatal sepsis. Therefore, we felt the need to modify our empiric antibiotic regimen, and decided to do so in a wellcontrolled manner. We, thus, compared our amoxicillincefotaxime regimen with an alternative penicillintobramycin regimen that is also appropriate therapy for this indication, but does not involve antibiotics that select for resistant gram-negative bacilli. This change in antibiotic policy was done as a prospective cross-over intervention trial in two identical units for neonatal intensive care.
Methods Participants and procedures
Department of Medical Microbiology and Infection Control, Erasmus University Medical Centre (P de Man MD, B A N Verhoeven MD, Prof H A Verbrugh MD, M C Vos MD); Department of Medical Microbiology and Infection Control, Sint Franciscus Gasthuis (P de Man); and Department of Paediatrics, Sophia’s Children’s Hospital, Erasmus University Medical Centre, Rotterdam, Netherlands (Prof J N van den Anker MD) Correspondence to: Dr P de Man, Department of Medical Microbiology and Infection Control, Sint Franciscus Gasthuis, PO Box 10900, 3004 BA, Rotterdam, Netherlands (e-mail: [email protected])
THE LANCET • Vol 355 • March 18, 2000
Our study was done in the division of neonatology, Sophia Children’s Hospital, which has three units. Two of these provide intensive care with mechanical ventilation. One of these two wards (unit A) has eight incubators whereas the other (unit B) has nine; otherwise the units are identical. Neonates no longer in need of this type of care are moved to a third unit without ventilator facilities. The two intensive units are adjacent to each other with a communicating door. Although unit A and B have separate teams of health-care workers, staff do sometimes use this door to work on the other unit. The effects of a change in antibiotic policy were studied as a prospective intervention, in which the two units were using different empiric antibiotic regimens. During the first 6 months of the study, unit A used intravenous amoxicillin combined with cefotaxime (amoxicillin-cefotaxime regimen) and unit B used
973
ARTICLES Penicillintobramycin regimen Number of admissions Measurements on admission Birthweight (g) Gestational age (weeks) Proportion delivered by caesarean section (%) Apgar at 1 min* (n=389) Apgar at 5 min† (n=390) Outcome measurements Weight at transfer (g) Time in unit (days) Number who died Proportion of days on ventilator (%) Proportion of days with central venous line (%)
218
Amoxicillincefotaxime regimen 218
1935 (1009) 33·3 (4·5) 42·4%
2101 (1104) 33·3 (4·5) 41·6%
5·6 (2·6) 7·7 (1·9)
6·0 (2·5) 7·9 (1·7)
2033 (921) 11·3 (16·5) 26 (11·9%) 40·4% 34·7%
p
2120 (980) 12·9 (20·5) 16 (7·3%) 40·0% 39·0%
0·2 0·6 0·9 0·9 0·23 0·4 0·001 0·14 0·7 0·005
Data are mean (SD) unless indicated. *Only 389 neonates were scored. †Only 390 neonates were scored.
Table 1: Characteristics of newborn babies penicillin G and tobramycin (penicillin-tobramycin regimen). After 6 months the units switched antibiotic regimens for a further 6 months (cross-over phase) to balance the effect of possible unit-related confounding variables. The chronic shortage of NICU beds in relation to the number of neonates made it impossible to randomly assign neonates to one of the treatment groups on admission. In practice, new neonates were admitted to the first available location, a vacancy often created by a transfer of another neonate to the non-ventilator unit. On the rare occasion where both ventilator units had capacity for new admission, the neonate was admitted to the unit with the lowest occupancy at that time. All new patients admitted to the two units were included in the analysis. Patients leaving the unit on the day of admission were not included in the analysis. The penicillin-tobramycin regimen involved these two antibiotics as empiric therapy for episodes of suspected septicaemia in the first 48 h of life (early onset). Flucloxacillin, which is also a narrow-spectrum penicillin, combined with tobramycin, was used for late-onset septicaemia (beyond the first 48 h of life). To avoid any use of amoxicillin and cephalosporins during the penicillin-tobramycin regimen, meropenem was used for culture-proven gram-negative bacteraemia. Meropenem was also prescribed for patients with suspected meningitis since there are no data for newborn babies on aminoglycoside penetration into the central nervous system. The regimen with broadspectrum penicillins and cephalosporins, the amoxicillincefotaxime regimen, used these two antibiotics for early-onset septicaemia. For late-onset septicaemia cefotaxime was combined with flucloxacillin. In month 4 of the study, after two cases of cefotaxime-resistant enterobacter bacteraemia had occurred, flucloxacillin was replaced by tobramycin for neonates with suspected late-onset septicaemia. Both regimens used vancomycin for persistent septicaemia in the presence of central venous lines. Neonates being treated with antibiotics at the time of crossover of the antibiotic regimens continued on their own regimens, but new courses of antibiotics were given in accordance with the prevailing regimen. In the analysis of the neonate-related measurements, such as Apgar scores and birthweight, the neonates were counted according to the regimen on admission. Antibiotic use was scored in relation to the regimen of the unit at that time. Both antibiotic regimens used in our study are routinely used in the Netherlands. We monitored a controlled change in antibiotic policy that did not involve investigational drugs or additional sampling. The study was discussed with the ethics committee. The committee considered this study to be a qualitycontrol investigation of the hospital, and not experimentation with human beings that would require formal ethics review and informed consent. However, parents and guardians were informed of all antibiotic treatments.
974
Colonisation strain
Penicillintobramycin regimen
Amoxicillincefotaxime regimen
Gram-negative bacillus resistant to cefotaxime Gram-negative bacillus resistant to tobramycin Gram-negative bacillus resistant to cefotaxime or tobramycin Cefotaxime-resistant Enterobacter spp Gram-negative bacillus resistant to empiric therapy of unit*
6·8 (16/2339)
21·4 (41/1914)
1·2 (3/2519)
0 (0/2706)
Relative risk (95% CI) 3·14 (1·76–5·56) ··
8·9 (19/2128)
21·4 (41/1914)
2·42 (1·41–4·15)
6·8 (15/2197)
20·3 (39/1917)
2·98 (1·64–5·38)
1·2 (3/2519)
21·4 (41/1914)
17·98 (5·57–58·01)
Data are (colonising events/patient days at risk) ⫻1000. *Tobramycin resistant in unit using penicillin-tobramycin regimen and cefotaxime resistant in unit using amoxicillin-cefotaxime regimen.
Table 2: Risk of colonisation with resistant gram-negative bacilli and antibiotic regimen Data on the neonates were collected on case record forms. For every neonate one form was used to record measurements such as Apgar scores and birthweight. Another case record form was used for each day of the admission with data such as antibiotic use, presence of central venous line, and type of ventilation for that particular day. Bacterial screening included cultures of respiratory aspirates and rectal swabs at admission and once a week thereafter. Samples were inoculated onto MacConkey and blood-agar plates. Plates were inspected after 18 h and 42 h of incubation at 37ºC. Each morphologically different colony type of potential pathogens, including all gram-negative bacilli, was selected for further identification. The number of different gram-negative bacteria is limited in these types and patients, usually there were only one or two different colony types and never more than four. When biochemistry and antibiograms of these colonies were identical, they were subsequently reported as one strain. Identification was carried out, and antibiograms were done with the MicroScan system (Dade International, West Sacramento, USA). Additional samples such as respiratory aspirates, blood, and cerebrospinal fluids were taken only when clinically indicated. Hygienic precautions included strict regulations for clothing, and hand disinfection with 70% ethanol was obligatory in between contact with each neonate.
Statistical analysis The data were analysed with the aid of the SAS system (version 6.12). Comparisons of means were done with the Student’s t test, and comparisons of proportion with Fisher’s exact test. The CIs for the relative risks were calculated with the Katz approximation. Colonisation with resistant bacteria was the primary outcome measurement of the effect of the change in antibiotic policy. The other comparisons were done to assess possible wanted and unwanted secondary effects of the changes in antibiotic regimen. Readers should be aware of the risk of significance by chance in these many comparisons.
Results From December, 1996, to December, 1997, 480 neonates were admitted to the two NICUs. 44 neonates left the unit on the day of admission and, therefore, were not included in this analysis. By chance, the remaining 436 admissions divided equally over the two regimens (218 admitted to each regimen). The neonates’ measurements on admission did not differ significantly between the two regimens (table 1). However, there were substantial differences between the two regimens with regard to the rate of colonisation with gram-negative bacilli. During the penicillin-tobramycin regimen E coli predominated and made up 53% of all gram-negative isolates. For the amoxicillin-cefotaxime regimen, 77% of all gram-negative isolates were Enterobacter spp. The species distribution of
THE LANCET • Vol 355 • March 18, 2000
ARTICLES
NICU A Isolates of bacilli resistant to empiric antibiotic regimen of unit Isolates of bacilli sensitive to empiric antibiotic regimen of unit but resistant to antibiotic regimen of other unit
1
4
8
12
16
20
24
28
32
Amoxicillin-cefotax ime regimen
36
40
44
48
52
48
52
Penicillin-tobramycin regimen
NICU B
1
4
8
12
16
20
24
28
Penicillin-tobramycin regimen
32
36
40
44
Amox icillin-cefotax ime regimen
Week number Isolation distribution of resistant bacilli from patients according to the antibiotic regimen
the remaining gram-negative isolates was not associated with the empiric antibiotic regimen. The risk of colonisation with resistant gram-negative bacilli expressed as colonising events per 1000 patient days at risk was calculated for five separate patterns of resistance. Emergence of resistance was higher during the amoxicillin-cefotaxime regimen (table 2). There were more episodes of colonisation with either tobramycinresistant or cefotaxime-resistant strains and there were more cefotaxime-resistant enterobacter among neonates on the amoxicillin-cefotaxime regimen. The most striking, and probably most relevant difference was found in the analysis of colonising events with strains resistant to the agents of the empirical antibiotic regimen in use on the unit. Colonising events with such strains occurred 18 times more frequently in the units when the amoxicillincefotaxime regimen was used than when the penicillintobramycin regimen was used (table 2). The development of resistance over time in the two units is shown in the figure. The colonisation with bacilli resistant to the empiric therapy is a reflection of antibiotic regimens, and closely follows the cross-over of antibiotic policy between the two NICUs. Colonisation with pathogens other than gram-negative bacilli, such as Candida spp, enterococci, Haemophilus spp, Listeria monocytogenes, coagulase-negative staphylococci, Staphylococcal aureus, and group-A and group-B streptococci, did not differ between the two
THE LANCET • Vol 355 • March 18, 2000
regimens (data not shown). This finding indicates that the problem with resistant gram-negative bacilli was not replaced by problems with other pathogens in the penicillin-tobramycin regimen. Despite the fact that 285 blood cultures were done within the first 48 h of life, there was a remarkable absence of culture-proven group-B streptococci bacteraemia. All cases of E coli bacteraemia occurred in neonates after day 6 of life. One neonate was colonised with L monocytogenes, which was not cultured from deep sites. Thus, there were no cases of early-onset bacterial neonatal sepsis or meningitis detected. However, there were nosocomial infections as complications of the Organism
Penicillintobramycin regimen
Amoxicillincefotaxime regimen
p
Candida spp Coagulase-negative staphylococci† Micrococcus luteus Staphylococcus aureus Enterococci Streptococci Bacillus cereus Enterobacter cloacae Klebsiella oxytoca Escherichia coli Total
2 15 0 0 0 2 0 0 1 7 27
1 23 2 1 2 1 1 5 0 1 37
. .* 0·23 ··* ··* ··* ··* ··* 0·06 ··* 0·07 0·22
Data are number of episodes of bacteraemia. *p values were not estimated for infections that occurred less than five times. †All but one of the coagulase-negative staphylococci were methicillin resistant in each regimen.
Table 3: Episodes of bacteraemia by causative organism
975
ARTICLES
Antibiotic
Penicillin-tobramycin regimen
Amoxicillin-cefotaxime regimen
Indication
% of Indication days* 28
Amoxicillin
Early-onset septicaemia Early and lateonset septicaemia None†
Cefotaxime
None†
Flucloxacillin
Late-onset septicaemia Culture proven gram-negative bacteraemia Late-onset septicaemia with central venous line
Penicillin Tobramycin
Meropenem
Vancomycin
34·7
None†
Late-onset septicaemia 1·2† Early-onset septicaemia 1·5† Early and lateonset septicaemia 9·8 Late-onset septicaemia 7·6 Cefotaxime-resistant gram-negative bacteraemia 3·2 Late-onset septicaemia with central venous line
p % of days* 0·8†
··
4
··
25·6
··
31·7
··
8·3
0·67
6·4
0·64
5·7
<0·001
*Number of days treated with antibiotic divided by total number of days multiplied by 100. †Used during the cross-over of regimens.
Table 4: Antibiotic use
intensive-care treatment. Differences in the rate of bacteraemia did not reach the 0·05 level of significance (table 3). However, enterobacter bacteraemia occurred five times in neonates on the amoxicillin-cefotaxime regimen but not at all in babies on the penicillintobramycin regimen (p=0·06). Conversely, E coli bacteraemia was seen more often with the penicillintobramycin regimen than with the amoxicillin-cefotaxime regimen (7 vs 1, p=0·07). There were no cases of tobramycin-resistant gramnegative bacteraemia among neonates on the penicillintobramycin regimen, indicating that the neonates on this regimen were always protected by these empiric antibiotics in case septicaemia occurred. There were four cases of cefotaxime-resistant gram-negative bacteraemia in neonates on the amoxicillin-cefotaxime regimen. After the first two cases occurred, the antibiotics for late-onset septicaemia used in this regimen were changed from flucloxacillin with cefotaxime to tobramycin with cefotaxime. Antibiotic use was a reflection of the empiric antibiotic regimens (table 4). The low-level use of penicillin during the amoxicillin-cefotaxime regimen and use of amoxicillin and cefotaxime during the penicillin-tobramycin regimen was a result of neonates continuing their course of antibiotics after the unit cross-over. There was one (short lived) violation of the protocol when a neonate in the penicillin-tobramycin regimen was given amoxicillin for an E coli infection. Remarkably, the use of meropenem, which was empiric therapy for meningitis and standard treatment for all cases of culture-proven gram-negative bacteraemia for neonates on the penicillin-tobramycin regimen, did not differ from the amoxicillin-cefotaxime regimen where it was only intended for use in cefotaximeresistant bacteraemia. Two outcome measurements differed significantly between the two regimens—the duration of stay in the unit was on average 1·5 days shorter for neonates treated with the penicillin-tobramycin regimen, and the use of central venous lines was lower for the penicillin-tobramycin regimen (table 1). This observation coincides with a higher vancomycin use in neonates on the amoxicillin-cefotaxime regimens. The other measurements did not differ significantly between the two regimens. The causes of death are listed in table 5. There was only one infection-related death, which occurred in a girl on chemotherapy for neuroblastoma. She had a central
976
Cause of death
Penicillintobramycin regimen
Amoxicillincefotaxime regimen
Asphyxia Encephalopathy Microencephalopathy Intracerebral bleeding Convulsions Twin to twin transfusion Neuroblastoma with Enterobacter sepsis Severe combined immunodeficiency syndrome Pulmonary hypoplasia Pulmonary bleeding Cystic lung malformation Pulmonary hypertension Meconium aspiration syndrome Hypoplastic left ventricle Tetralogy of Fallot Pneumopericardium Triploidy Pyruvate dehydrogenase deficiency Unknown Total
5 2 1 4 1 2 0 0 6 0 0 1 1 1 1 0 1 0 0 26*
4 0 0 0 0 0 1 1 3 1 1 0 0 1 1 1 0 1 1 16†
*There were no deaths attributed to renal failure. †There were no significant differences.
Table 5: Causes of death
venous line and developed sepsis with cefotaximeresistant enterobacter. She remained line dependent because of septic shock. Although the enterobacter was sensitive towards meropenem and tobramycin, treatment with these drugs could not reverse the bacteraemic state in the presence of the line. The other fatal cases occurred in patients with major underlying diseases or abnormalities. Careful re-assessment of the patients’ files did not show a role for infection in those cases. The amoxicillin-cefotaxime regimen was designed to use cefotaxime with flucloxacillin for late-onset septicaemia. However, in the fourth month of the study two neonates that developed this disease entity were treated with meropenem because they were known to be colonised with cefotaxime-resistant enterobacter. They recovered, and indeed their blood cultures were positive with these resistant bacteria. After these incidents, the study protocol was changed to contain tobramycin for all neonates developing late-onset septicaemia, since colonisation cultures do not always precede sepsis and we wanted to guarantee the best treatment for all children. These interventions however, may have concealed the effect of chance in antibiotic policy on mortality, since they may well have caused a decrease in mortality in the amoxicillin-cefotaxime regimen.
Discussion This study shows that policies about the empiric use of antibiotic regimens do matter in the control of antimicrobial resistance in an intensive-care setting. The flora of the body surfaces forms the reservoir for invasive infections. We found a 18-fold higher rate of colonisation with strains resistant to the empiric therapy in patients on the amoxicillin-cefotaxime regimen than in those on penicillin-tobramycin regimen. The newborn babies treated with the penicillin-tobramycin regimen were thus better protected against nosocomial septicaemia with colonising gram-negative strains than those who were routinely treated with amoxicillin and cefotaxime. In this paper we do not describe an analysis of the factors associated with the risk for colonisation with resistant bacteria in these neonates, other than the antibiotics used. We did do such an analysis, which showed that risk for colonisation was related to low
THE LANCET • Vol 355 • March 18, 2000
ARTICLES
gestational age, low birthweight, and the length of stay, factors known from previous studies. Numerous other factors did not contribute. The analysis was done as a case-control study, considerably different from the cohort study presented in this paper, and therefore will be published elsewhere. The number of prospective studies correlating antibiotic regimens and bacterial resistance is limited, and there is no absolute proof of a causative association between antibiotic use and resistance.13 The most relevant information on this subject originates from reports on the management of outbreaks and other studies with historical controls. Price and Sleigh14 halted an outbreak of resistant klebsiella by a complete cessation of all antibiotic prescribing in a neurosurgical intensive-care unit. In a preintervention versus postintervention study in a paediatric intensive-care unit, Toltzis and co-workers15 were able to reduce ceftazidime use by 96% but were confronted with an increase in ceftazidime resistance. However, White and colleagues16 restricted antibiotic use in the management of an outbreak of resistant acinetobacter and achieved a significant decrease not only in resistant acinetobacter but also in the antibiotic resistance of Enterobacteriaceae. In Finland a nationwide reduction of macrolide use was associated with a reduction in macrolide resistance among group-A haemolytic streptococci.17 The antibiotics in the traditional regimens in NICUs have been selected for their activity against early-onset neonatal septicaemia with pathogens acquired in utero or during labour. However, there have been changes in NICUs over the years. The admissions have shifted towards more premature infants often delivered by caesarean section for maternal reasons not related to infection. These premature infants are in need of intensive care over extended periods of time. They are seldom born with infection but are at increasing risk of acquiring infection during their stay in the NICU.18 A remarkable observation in this study was the absence of typical earlyonset neonatal septicaemia. An antibiotic regimen should not only treat the normal pathogens of neonatal septicaemia but it should also provide protection against the nosocomial pathogens that threaten the newborn babies during their stay in the NICU. Antibiotic treatment by the classic antibiotic regimens in NICU, tends to select bacteria with resistance towards the antibiotics used. In the amoxicillin-aminoglycoside regimens there is a selective pressure towards -lactamase producers that can degrade the amoxicillin. Such strains including klebsiellae have a clear advantage over normally sensitive bacteria such as E coli, and are, therefore, likely to become resident ward strains. These ward strains are then repeatedly exposed to the aminoglycosides and in turn may become aminoglycoside-resistant strains with outbreak potential.2–7 Cephalosporins evade the problems of resistance because they have less effect on the normal intestinal flora.19 This theory is supported by the observation that neonates treated with amoxicillin had more resistant gram-negative bacilli than neonates who had been on cefuroxime treatment.20–23 In the cefotaxime-containing regimens used in NICUs including ours, cefotaxime was combined with amoxicillin to cover Listeria spp. In these circumstances the amoxicillin will eliminate the normal intestinal flora. Gram-negative bacteria such as Enterobacter spp and Serratia spp that can degrade
THE LANCET • Vol 355 • March 18, 2000
cefotaxime and amoxicillin will be selected and may flourish.7,24,25 Resistant enterobacter strains will cause invasive infections in NICU and cannot be treated with cefotaxime.8 The resistance of enterobacter has been seen to reach a plateau at 25–35% resistant strains. This is explained by the fact that most enterobacter infections arise from the patients own endogenous flora. This argument clearly does not hold for NICU, where the ward flora is the reservoir for both colonisation and infection. The penicillin-tobramycin regimen differs from these two regimens by the fact that it does not include a lactam with a selective effect towards resistant gramnegative bacilli. Penicillin G does not penetrate the outer membrane of gram-negative bacilli and, therefore, lacks activity towards these bacteria. Thus, it does not generate a selective pressure towards resistant, -lactamaseproducing bacilli. Meropenem was the only broadspectrum -lactam we used in the penicillin-tobramycin regimen. It was chosen as a treatment for culture positive gram-negative bacteraemia, because it lacks a selective effect towards bacteria such as Enterobacter spp and Serratia spp—largely due to its stability towards most lactamases. However, an overuse of this drug can evoke problems with strains of Stenothrophomonas maltophilia, which produces -lactamase capable of inactivating carbapenems. Use of aminoglycoside in the penicillin-tobramycin regimens was not associated with a significant emergence of resistance during the study (only three colonising events in 2515 patient days). Aminoglycosides, given parenterally, do not penetrate the mucosal surfaces of the gastrointestinal tract and therefore may have a low selective effect. In fact, our study and others26 show that newborn babies on aminoglycoside therapy become colonised with gram-negative bacteria sensitive to these antibiotics. The choice of -lactams that have no selective effects—either alone or in a combination with an aminoglycoside that does not penetrate mucosal membranes—may be the reason for the striking reduction in resistance seen in this study. Penicillin G and tobramycin do cover the pathogens of early-onset neonatal septicaemia, since almost all community acquired E coli are sensitive to tobramycin and group-B streptococci and Listeria spp are sensitive to penicillin G. The empiric treatment of suspected bacterial meningitis remains a problem, since the penetration of aminoglycosides into the central nervous system is poor in adults. The undeveloped blood-brain barrier in preterm infants might allow for more effective penetration. Since we are not aware of studies on this subject, we used meropenem when meningitis was suspected. The high amount of antibiotic use, in combination with the low grade of colonisation of neonates at the time of their admissions, turns the NICU into an environment where antibiotic policy is likely to have a pronounced effect on the resistance problem. These facts contribute to the clear differences found in this study. Although in other wards the effects of antibiotic therapy can be less obvious, the mechanisms behind the selection of resistant nosocomial strains are probably similar if not identical. In fact, Enterobacter spp and other -lactamase producing bacteria are frequently encountered throughout the hospital.27 The approach of this change in policy, avoiding broad-spectrum penicillins and cephalosporins could be used in other wards that have problems with resistant Enterobacteriaceae, or such a change can be seen as a
977
ARTICLES
preventive measure that will be of value in wards that use high quantities of antibiotics. We have been using the penicillin-tobramycin regimen for neonatal sepsis in all our NICUs for 2 years since the end of the study. There have been no further or new problems with resistance or treatment failures in this period. This study provides solid evidence of a reduction of bacterial resistance mediated by the use of antibiotics that exert little selective pressure on the mucosal environment, confirming the benefits of the often preached but less frequently practised use of narrow-spectrum antibiotics. This study might inspire clinicians to put that old knowledge into practice more often.
9
Contributors
14
P de Man did the initial protocol design, coordinated the microbiology, and contributed to the statistical analysis and the writing of the paper. B A N Verhoeven coordinated and collected clinical data and contributed to the statistical analysis and writing of the paper. H A Verbrugh contributed to the protocol design, the analysis, and the writing of the paper. M C Vos contributed to the writing of the paper. J N van den Anker coordinated the clinical aspects of the study, reviewed the charts of all fatalities, designed and analysed the study, and wrote the paper.
10
11
12
13
15
16
Acknowledgments We thank the nursing staff of the department of neonatology, the laboratory technicians, and A Smit of the department of medical microbiology and infectious diseases for their conscientious work during this study; D van Dongen for collection of clinical data; M Vogel for extracting computer information from the hospital system; J G M Koeleman, J W Mouton, C Verduin, and I C Gyssens for discussion and proofreading; and D M MacLaren for his original idea to avoid all selective effects of -lactams by the use of penicillin G instead of amoxicillin in the control of a serratia outbreak in 1994 (NICU of the Free University, Amsterdam, Netherlands).
17
References
20
1
2
3
4
5
6
7
8
Hammerschlag MR, Klein JO, Herschel M, Chen FC, Fermin R. Patterns of use of antibiotics in two newborn nurseries. N Engl J Med 1977; 296: 1268–69. Noy JH, Ayliffe GA, Linton KB. Antibiotic-resistant gram-negative bacilli in the faeces of neonates. J Med Microbiol 1974; 7: 509–20. Mayhall CG, Lamb VA, Bitar CM, et al. Nosocomial klebsiella infection in a neonatal unit: identification of risk factors for gastrointestinal colonization. Infect Control 1980; 1: 239–46. Markowitz SM, Veazey JM Jr, Macrina FL, Mayhall CG, Lamb VA. Sequential outbreaks of infection due to Klebsiella pneumoniae in a neonatal intensive care unit: implication of a conjugative R plasmid. J Infect Dis 1980; 142: 106–12. Kayyali MZ, Nicholson DP, Smith IM. A klebsiella outbreak in a pediatric nursery: emergency action and preventive surveillance. Clin Pediatr (Phila) 1972; 11: 422–26. Arredondo-Garcia JL, Diaz-Ramos R, Solorzano-Santos F, Sosa-Gonzalez IE, Beltran-Zuniga M. Septicaemia due to Klebsiella pneumoniae in newborn infants: nosocomial outbreak in an intensive care unit. Rev Latinoam Microbiol 1992; 34: 11–16. Bryan CS, John JF Jr, Pai MS, Austin TL. Gentamicin vs cefotaxime for therapy of neonatal sepsis: relationship to drug resistance. Am J Dis Child 1985; 139: 1086–89. Modi N, Damjanovic V, Cooke RW. Outbreak of cephalosporin resistant Enterobacter cloacae infection in a neonatal intensive care unit. Arch Dis Child 1987; 62: 148–51.
978
18
19
21
22
23
24
25
26
27
Haertl R, Bandlow G. Molecular typing of Enterobacter cloacae by pulsed-field gel electrophoresis of genomic restriction fragments. J Hosp Infect 1993; 25: 109–16. Spritzer R, Kamp HJ, Dzoljic G, Sauer PJ. Five years of cefotaxime use in a neonatal intensive care unit. Pediatr Infect Dis J 1990; 9: 92–96. Verweij PE, Van Belkum A, Melchers WJ, Voss A, HoogkampKorstanje JA, Meis JF. Interrepeat fingerprinting of third-generation cephalosporin-resistant Enterobacter cloacae isolated during an outbreak in a neonatal intensive care unit. Infect Control Hosp Epidemiol 1995; 16: 25–29. Venezia RA, Scarano FJ, Preston KE, et al. Molecular epidemiology of an SHV-5 extended-spectrum beta-lactamase in enterobacteriaceae isolated from infants in a neonatal intensive care unit. Clin Infect Dis 1995; 21: 915–23. Gould IM. A review of the role of antibiotic policies in the control of antibiotic resistance. J Antimicro Chemoth 1999; 43: 459–65. Price DJ, Sleigh JD. Control of infection due to Klebsiella aerogenes in a neurosurgical unit by withdrawal of all antibiotics. Lancet 1970; ii: 1213–15. Toltzis P, Yamashita T, Vilt L, et al. Antibiotic restriction does not alter endemic colonization with resistant gram-negative rods in a pediatric intensive care unit. Crit Care Med 1998; 26: 1893–99. White AC Jr, Atmar RL, Wilson J, Cate TR, Stager CE, Greenberg SB. Effects of requiring prior authorization for selected antimicrobials: expenditures, susceptibilities, and clinical outcomes. Clin Infect Dis 1997; 25: 230–39. Seppala H, Klaukka T, Vuopio-Varkila J, et al. The effect of changes in the consumption of macrolide antibiotics on erythromycin resistance in group A streptococci in Finland. Finnish Study Group for Antimicrobial Resistance. N Engl J Med 1997; 337: 441–46. Khadilkar V, Tudehope D, Fraser S. A prospective study of nosocomial infection in a neonatal intensive care unit. J Paediatr Child Health 1995; 31: 387–91. Ledingham IM, Alcock SR, Eastaway AT, McDonald JC, McKay IC, Ramsay G. Triple regimen of selective decontamination of the digestive tract, systemic cefotaxime, and microbiological surveillance for prevention of acquired infection in intensive care. Lancet 1988; i: 785–90. Tullus K, Burman LG. Ecological impact of ampicillin and cefuroxime in neonatal units. Lancet 1989; i: 1405–07. Tullus K, Berglund B, Berman LG. Emergence of cross-resistance to beta-lactam antibiotics in fecal Escherichia coli and Klebsiella strains from neonates treated with ampicillin or cefuroxime. Antimicrob Agents Chemother 1990; 34: 361–62. Burman LG, Berglund B, Huovinen P, Tullus K. Effect of ampicillin versus cefuroxime on the emergence of beta-lactam resistance in faecal Enterobacter cloacae isolates from neonates. J Antimicrob Chemother 1993; 31: 111–16. Kalenic S, Francetic I, Polak J, Zele-Starcevic L, Bencic Z. Impact of ampicillin and cefuroxime on bacterial colonization and infection in patients on a neonatal intensive care unit. J Hosp Infect 1993; 23: 35–41. Acolet D, Ahmet Z, Houang E, Hurley R, Kaufmann ME. Enterobacter cloacae in a neonatal intensive care unit: account of an outbreak and its relationship to use of third generation cephalosporins. J Hosp Infect 1994; 28: 273–86. Berkowitz FE, Metchock B.Third generation cephalosporin-resistant gram-negative bacilli in the feces of hospitalized children. Pediatr Infect Dis J 1995; 14: 97–100. Blakey JL, Lubitz L, Barnes GL, Bishop RF, Campbell NT, Gillam GL. Development of gut colonisation in pre-term neonates. J Med Microbiol 1982; 15: 519–29. Sanders WE Jr, Sanders CC. Enterobacter spp: pathogens poised to flourish at the turn of the century. Clin Microbiol Rev 1997; 10: 220–41.
THE LANCET • Vol 355 • March 18, 2000