Corresponding author and reprint requests: Ø. Wendelbo, Department of Medicine, Haukeland University Hospital, N-5021 Bergen, Norway Tel: +47 55 97 50 00 Fax: +47 55 97 29 50 E-mail: firstname.lastname@example.org
Objectives To examine and characterize a suspected outbreak of high-level gentamicin-resistant Enterococcus (HLGRE) infection.
Methods Eighty-nine patients with clinical infection diagnosed during hospital stay or within 30 days after discharge in the period from June 1995 to 31 December 1999 were included in the study. One control patient was assigned for each HLGRE patient according to localization in the hospital (same ward), time of admission (±3 months), and age (±10 years). Unadjusted risk analysis and multivariate logistic regression analysis were performed. Sixty-nine HLGRE strains were subjected to PCR amplification of the genes coding for aminoglycoside-3′-O-phosphoryltransferase-III (APH(3′)-III) and aminoglycoside-6′-N-acetyltransferase/2′′-O-phosphoryltransferase-III (AAC(6′)/APH(2′′)).
Results The gene aacA/aphD, associated with HLGRE, was detected by PCR in all isolates, and the gene aphA3, associated with high-level streptomycin, kanamycin and amikacin resistance, was detected in 56 of the 69 isolates. None of the 69 isolates was resistant to glycopeptides or ampicillin. Resistance to ciprofloxacin was found in 57 (82.6%). Pulsed-field gel electrophoresis analysis revealed 12 different genotypes, among which two major clusters dominated.
Conclusions Both clonal expansion and the emergence of unique strains contributed to the increased number of infections caused by HLGRE. Urinary catheterization, duration of hospital stay and antibiotic therapy were significant risk factors for HLGRE infection.
Enterococcus faecalis is intrinsically resistant to a number of antimicrobial agents. It is tolerant to penicillins, and can acquire high-level resistance to penicillins through the production of altered penicillin-binding proteins (PBPs) [1,2] or, very rarely, through the production of β-lactamase . It displays resistance to cephalosporins and low-level resistance to clindamycin, aminoglycosides and fluoroquinolones, and is frequently resistant to chloramphenicol, macrolides, and tetracyclines.
High-level gentamicin-resistant Enterococcus (HLGRE) was first reported in France in 1979 , and then in the USA in 1983 . Since then, increasing numbers of HLGRE isolates have been reported from several continents [6–12].
The majority of infections caused by enterococci can be cured if the patient has a normal host defense. For the successful treatment of endocarditis and other serious enterococcal infections, a bactericidal combination of aminoglycosides and cell wall-active agents is required. When enterococci have acquired high-level resistance to gentamicin (minimum inhibitory concentration (MIC) ≥ 512 mg/L), the synergistic effect of aminoglycosides and β-lactam antibiotics is not obtained .
High-level gentamicin resistance in enterococci is mainly mediated by the bifunctional aminoglycoside-modifying enzyme, aminoglycoside-6′-N-acetyltransferase/2″-phosphotransferase-III (AAC(6′)/APH(2″)), encoded by the gene aacA/aphD. This enzyme has dual activity, conferring high-level resistance to gentamicin and to all available aminoglycosides  except streptomycin and neomycin. High-level streptomycin resistance (MIC ≥ 1024 mg/L) is ribosomally mediated , due to an alteration in a single protein, S12, in the 30S ribosomal subunit , or by the synthesis of the aminoglycoside adenyltransferase AAD (6′) . Previous reports have documented an increasing incidence of HLGRE in hospital settings , and intra-hospital clonal outbreaks have also been described .
The object of this study was to examine an outbreak caused by HLGRE in our hospital and to characterize this outbreak microbiologically and epidemiologically. We also wanted to study the antibiotic resistance pattern and mechanisms of gentamicin resistance of the isolates, and to elucidate risk factors for infection with HLGRE.
Materials and methods
Haukeland University Hospital is a fully specialized 1100-bed hospital serving as a referral hospital and an emergency hospital for populations of 1 million and 350 000, respectively. Until 1997, HLGRE was infrequently diagnosed; however, it was not specifically tested for. Later, the mean isolation rate of HLGRE increased to 13% of all enterococcal isolates.
Definition of cases and controls
All patients with a clinical HLGRE infection diagnosed during their hospital stay or within 30 days after discharge in the period 1995–2000 were included in the study. By means of the hospital computer system, one control patient was identified for each case according to localization in the hospital (same ward), time of admission (±3 months), and age (±10 years).
Collection of clinical data
Data were collected from the patient records. The variables included age, gender, prior hospital admissions, number of days in the hospital, antimicrobial agents received, days on antimicrobial treatment, underlying diseases, main diagnosis, use of intravenous catheters, use of urinary catheters, immunosuppressive therapy, ward, infection site, surgery performed, and outcome. Underlying disease was classified according to the tenth revision of the International Classification of Diseases (ICD-10). Previous hospitalization with discharge within the last month was added to the length of the present hospital stay. Antibiotics prescribed within six months prior to the hospitalization were added to the total consumption of antibiotics.
The paired Wilcoxon test was applied for comparing data from continuous variables between cases and controls. For each dichotomous variable, the matched 2 × 2 table for cases versus controls was analyzed by calculating the odds ratio (OR), applying McNemar's test. A 95% confidence interval (CI) was calculated with exact simple conditional logistic regression (LR). When the ordinary OR estimate was impossible to calculate, the median unbiased estimate (MUE) was used . In this case, the 95% CI is more reliable . In the order of their statistical significance determined from McNemar's test, variables were included in a forward stepwise exact conditional analysis. A variable was retained in the model for the next step if its exact likelihood ratio test had a P-value <0.05. From the final model, adjusted ORs with 95% CIs were estimated, and P-values were calculated from the LR test. Basic calculations were performed by SPSS version 9.0 (SPSS Inc., Chicago, IL, USA). Exact logistic regression analysis was performed with Log Xact 4.1 (Cytel Software Corporation, Cambridge, MA, USA).
Bacterial isolates and antibiotic susceptibility testing
Enterococcal isolates were identified by standard biochemical methods . The β-lactamase activity of all isolates was tested with nitrocefin paper disks (AB Biodisk, Solna, Sweden). The susceptibilities of the isolates to antimicrobial agents were examined by an agar diffusion method , using paper disks and PDM Antibiotic Sensitivity Medium from AB Biodisk. The breakpoint for resistance to ampicillin is ≥ 32 mg/L, and that for gentamicin is ≥ 8 mg/L, according to the recommendations given by The Norwegian Working Group on Antibiotics . The Etest (AB Biodisk) was used to examine all isolates classified as resistant to gentamicin for resistance to ciprofloxacin, teicoplanin, and vancomycin, and high-level resistance to gentamicin or streptomycin. An isolate was designated as having high-level resistance to gentamicin if the MIC was ≥ 512 mg/L. These isolates were examined by polymerase chain reaction (PCR) for verification of the presence of aacA/aphD.
Pulsed-field gel electrophoresis
Pulsed-field gel electrophoresis (PFGE) was performed on SmaI-digested (Promega Corp., Madison, WI, USA) genomic DNA as previously described , with modifications according to Dahl et al . The PFGE patterns were interpreted as previously described by Tenover et al. . Isolates were considered to be indistinguishable if their restriction patterns had the same number of bands and the corresponding bands had the same apparent molecular size. An isolate was considered to be closely related to the outbreak strain and probably part of the outbreak if its PFGE pattern differed by two or three bands, consistent with a single genetic event. Isolates that differed by four to six bands, consistent with two genetic events, were considered to be possibly related. Strains that differed by seven bands or more from the outbreak strain were considered not to be part of the outbreak.
Sixty-nine HLGRE strains were subjected to PCR amplification of the genes coding for aminoglycoside-3′-O-phosphoryltransferase-III (APH(3′)-III) and aminoglycoside-6′-N-acetyltransferase/2′′-O-phosphoryltransferase-III (AAC(6′)/APH(2′′)), as previously described , with the following modification: one colony dispersed in 100 µL of TE (0.01 m Tris-HCl, 0.001 m EDTA) was boiled for 10 min, and centrifuged at 5000g for 5 min, and 1 µL of the supernatant was used as template for the PCR reaction.
Between 8 June 1995 and 31 December 1999, 93 patients with a clinical infection with HLGRE were identified. Twenty-four bacterial isolates were not available for further analysis, leaving 69 bacterial isolates from 69 patients for further microbiological and genetic characterization. Clinical information was available from the medical records of 89 patients. Four patients were not included in the case–control study, due to incomplete medical data. Only the 89 patients with complete data were matched with controls.
The first two cases of HLGRE infection were detected at Haukeland University Hospital in the first half of 1995. They were hospitalized in the intensive care unit (ICU), burns unit, and medical ward. The epidemic reached a peak in the first half of 1999, and showed a slight decrease in the second half of 1999 (Figure 1). The numbers of HLGRE isolates obtained in the first few months of 2000 indicated a further increase in HLGRE infections (data not shown). We do not know the reason for the slight decrease in the second half of 1999, since there were no extra infection control measures. Of the 93 patients included in the study, 33 were hospitalized in the Department of Surgery, 19 in the Department of Medicine, and seven in the Department of Oncology. Departments with four or less cases were designated ‘other’, and accounted for 34 of 93 isolates (Figure 2). The incidence of HLGRE was highest in the Department of Surgery, with 193 cases/100 000 patient-days, followed by the ICU, with 60 cases/100 000 patient-days, and the burn unit, with 52.2 cases/100 000 patient-days (data not shown). The first isolate was not available for PFGE, and the next three were not genetically related.
Microbiological and genetic characterization
None of the 69 bacterial isolates was resistant to glycopeptides or ampicillin. Resistance to ciprofloxacin was found in 57 of 69 (82.6%). All the resistant isolates had a ciprofloxacin MIC ≥ 32 mg/L, and all the sensitive isolates had a ciprofloxacin MIC <4 mg/L. The range of ciprofloxacin MICs (mg/L) was 0.038 to ≥ 32 mg/L. High-level resistance to streptomycin was detected in 56 of the 69 (81.2%) isolates. The gene aacA/aphD encoding the bifunctional enzyme AAC(6′)-APH(2′) was detected by PCR in all 69 isolates, and the gene aphA3 encoding the APH(3′)-III protein was found by PCR in 58 of the 69 isolates (84.1%). PFGE analysis revealed 12 different genotypes, among which two major clusters dominated, designated cluster A (n = 48, 69.6%) and cluster B (n = 8, 11.6%). There were no differences in antibiotic profile between cluster A and cluster B.
Distribution of specimen and case–control matching
Forty-seven (52.8%) of the HLGRE-infected patients had urinary tract infection, 29 (32.6%) wound infection, four (4.5%) bacteremia, and nine (10.1%) other infections. Matching controls with complete data were successfully found for all the 89 included patients. There were 36 females and 53 males in the case group, and 39 females and 50 males in the control group. The baseline characteristics of these patients are shown in Table 1. The mean age was 64.9 years (range 0–91 years) for the cases, and 66.4 years (range 0–90 years) for the controls (Table 1).
Table 1. Epidemiologic data for 89 patients infected with high-level gentamicin-resistant enterococci and 89 control patients
HLGRE patients N = 89
Control patients N = 89
Paired Wilcoxon test except for percentage of females (McNemar's test).
Unadjusted analyses of possible risk factors for HLGRE infection
The duration of the hospital stay was significantly longer for the cases than for the controls (median 33.0 versus 9.0 days) (Table 1). A cut-off value of 14 days was subsequently used for dichotomization. Table 2 gives an overview of the significant risk factors according to the univariate analysis. Surgical procedures were performed significantly more frequently in the cases than in the controls (50.6% versus 32.6%). There was a positive association between infection with HLGRE and urinary catheterization, which was not the case for intravascular catheters. Exposure to antibiotics was a significant risk factor for acquiring HLGRE infection (OR = 3.5), as well as specific exposure to macrolides (OR = 9.6), quinolones (OR = 5.8), aminoglycosides (OR = 5.3), metronidazole (OR = 5.0), second-generation cephalosporins (OR = 4.3), isoxazolylpenicillin (OR = 4.0), penicillin G (OR = 3.1), and third-generation cephalosporins (OR = 2.7).
Table 2. Unadjusted analysis of risk factors for HLGRE infections
Percentage of patients exposed
Cases (n = 89)
Controls (n = 89)
a MUE, median unibased estimate, ‘Rely on CI ‘(see Appendix A.6 in CYTEL Software Corporation ).
Intravenous catheters, immunosuppressive therapy, underlying disease, intra-hospital death, gender and prescription of first-generation cephalosporins, co-trimoxazole, vancomycin, teicoplanin, trimethoprim, clindamycin, carbapenems, mecillinam, tetracyclines and nitrofurantoin were included in the univariate analysis, but were not statistically significant risk factors.
Multivariate analysis of possible risk factors for HLGRE infection
In the forward stepwise LR analysis that was conducted using the significant risk factors from the univariate analysis, prescription of macrolides and quinolones, number of days in hospital (≤14 versus ≥ 15 days), number of days on antibiotics (≤14 versus ≥ 15 days), urinary catheterization, prescription of penicillin, second- or third-generation cephalosporins, metronidazole and isoxazolylpenicillin, surgery conducted and prescription of any antibiotic were included in the model. Number of days in hospital, number of days on antimicrobial treatment and urinary catheterization appeared as significant risk factors in the multivariate analysis.
The intra-hospital death rate was 5.6% for the cases, compared to 2.3% for the controls, corresponding to a 3.4% excess mortality in the case group. This difference was not statistically significant (P = 0.08). We did not find HLGRE infection to be the major cause of death for any of the patients. There was no significant difference between the cases and the controls regarding underlying disease (data not shown).
Antibiotic-resistant enterococci have so far not been a problem in Scandinavia, although some studies have reported both ampicillin-resistant enterococci and vancomycin-resistant enterococci [28–31]. In total, five infections caused by vancomycin-resistant enterococci have been reported in our hospital. Regarding HLGRE in Norway, only one study exists, reporting 6% of fecal carriers in a large teaching hospital .
The nosocomial outbreak at Haukeland Hospital described in the present study was dominated by two genotypes, indicating a clonal expansion. We did not find any obvious differences between patients infected with cluster A and those infected with cluster B. Explorations of the data do not allow any conclusion as to how the resistant strains were introduced into the hospital. Nosocomial spread from patient to patient has been well documented . Zervos et al.  were able to isolate resistant enterococci from the environment and from the hands of healthcare personnel. They concluded that transient carriage of the organism on hands might have been the mode of transmission.
Several strains not genetically related to the outbreak strain contributed to the total burden of HLGRE in the present study. These strains may have formed part of the patient's bowel flora, and may have been introduced as the patient entered the hospital. High-level gentamicin resistance in all the 69 E. faecalis isolates was due to the presence of the aminoglycoside-modifying enzyme AAC(6′)/APH(2′′). The gene aacA/aphD has in most cases been associated with plasmids. The occurrence of several unrelated strains in addition to the outbreak strain, all with the gene aacA/aphD, may indicate horizontal spread of genetic elements containing this gene. Even though we have no direct evidence for person-to-person spread, or spread of genetic elements, this has previously been shown , and remains a possibility for the mode of spread in the present outbreak.
Previous studies have documented that enterococci cause 5–20% of all endocarditis cases [36–38]. None of the patients included in the present study suffered from endocarditis. Underlying disease has been demonstrated as a significant risk factor for acquiring infection with HLGRE . Neither the univariate analysis nor the multivariate analysis in our study confirmed such a hypothesis.
Over the last two decades, several risk factors for acquiring infections with HLGRE have been identified. Various studies have reported prior prolonged antibiotic treatment, number of antimicrobial agents prescribed, prior surgical procedures, perioperative antibiotic prophylaxis, duration of hospitalization and antibiotic therapy, instrumentation, and renal failure [34,40–42]. Some of these are linked to each other, and it has been unclear which of them are truly important. For instance, prescription of certain broad-spectrum antimicrobial agents may reflect severe infections that have not responded to other treatment, and that consequently lead to an increase in the number of antimicrobial agents prescribed and the number of days on antimicrobial treatment.
The present study included patients who had clinical infections with strains of E. faecalis highly resistant to gentamicin. The multivariate analysis showed that patients who stayed for more than 14 days in hospital, who had been catheterized in the urinary tract and who had received antibiotics for more than 14 days were most likely to be infected with HLGRE. Patients who had been hospitalized for more than two weeks were about six times more likely to be infected with HLGRE than patients who had been hospitalized for shorter periods. These results are in accordance with previous reports .
Cephalosporin, quinolone and aminoglycoside use have been documented to be independent risk factors [9,40]. In the present study, quinolone, metronidazole, second- and third-generation cephalosporin, aminoglycoside, isoxazolylpenicillin and macrolide prescription were significant risk factors according to the univariate analysis. These results were not confirmed by multivariate analysis. The fact that enterococci are resistant to several antibiotics, and several of the agents are active in the gut, and thereby affect the microbial gut flora, may be the reason why the use of macrolides or quinolones was found to be significant in the univariate analysis. However, these findings were not supported by the results of the multivariate analysis.
Prescription of cephalosporins has increased moderately in Norway over the last five years (0.52 DDD/1000 inhabitants per day) , and the average consumption of cephalosporins at Haukeland hospital is similar to that in other hospitals in Norway (data not shown).
The prescription of cephalosporins is still low in Norway compared to other countries in Europe and the USA. This relatively low consumption of cephalosporins might explain why their usage was not a significant risk factor in the final, adjusted analysis.
The finding of prolonged hospital stay, prolonged antibiotic treatment and urinary catheterization as significant risk factors for HLGRE infection illustrates the impact of both long exposure and risk procedures, and is in concordance with the results from other studies [42,44]. When patients become infected with HLGRE, the therapeutic options are limited. It is therefore important to protect patients at risk from developing serious infections such as endocarditis and osteomyelitis. In order to do so, effective infection control is necessary to limit the spread of these resistant bacteria within the hospital. However, it is hardly realistic to target risk patients infected with HLGRE, and therefore we believe in general precautions as the most effective infection control measure.
We conclude that HLGRE infection is a significant clinical problem at our hospital, and that both clonal expansion and the emergence of unique strains have contributed to the increase in infections caused by HLGRE. The number of days in hospital, duration of antibiotic treatment and urinary catheterization are significant risk factors for acquiring infections with HLGRE. Although we consider HLGRE to represent a serious risk for certain groups of patients, we were not able to show increased intra-hospital death rates among the patients.