INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Nosocomial infections caused by Enterococcus spp. are growing in importance. In Spain, nosocomial infections produced by these organisms constitute 9.4% of all infections, and nosocomial enterococcal infections are fourth in frequency after those caused by Escherichia coli, Pseudomonas aeruginosa, and Staphylococcus aureus (31). The frequency of nosocomial enterococcal bacteremia rose from 4.6% in 1990 to 7.1% in 1994, and 70.4% of the cases of bacteremia were caused by Enterococcus faecalis (31). According to data from the National Nosocomial Infections System of the Centers for Disease Control and Prevention, in the United States Enterococcus spp. are responsible for 12% of all nosocomial infections and 8% of all bacteremias acquired nosocomially (30).

Enterococcus spp. have a great capacity for acquiring resistance to antimicrobial agents. The resistance to penicillins or vancomycin and high-level resistance to aminoglycosides are the most important. For most enterococcal infections, the use of a penicillin (or a glycopeptide agent if the organism is resistant to penicillin or the patient is allergic to penicillin has been shown to be sufficient treatment (19). In patients with severe infections, in which it is necessary to obtain bactericidal activity at the site of the infection, the use of a combination of an antimicrobial agent with activity against the bacterial cell wall (penicillin or glycopeptide) and an aminoglycoside is recommended (18, 19). When the strain causing the infection presents high-level resistance to the aminoglycosides, use of the latter combination is no longer possible. Therefore, the study of the characteristics of infections caused by enterococcal strains with this phenotype is of interest, because the management of or the prognosis for patients infected with such strains may need to be modified. The objective of the present study was to compare the epidemiological, microbiological, clinical, and prognostic characteristics of bacteremias caused by Enterococcus spp. with and without high-level resistance to gentamicin (HLRG).

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Design of the study. A prospective, multicenter study was carried out. Thirteen community and university hospitals in Andalusia, Spain, participated in the study from June 1993 to March 1994. All patients whose blood was found to be positive for Enterococcus spp. on culture were included. Those patients whose positive blood cultures were not considered to be of clinical significance were excluded. A clinical follow-up until the discharge or the death of the patient was carried out for each patient. The following information was collected for each patient: epidemiological data (personal data, age, sex, hospitalization ward, means of acquisition, risk factors for the acquisition of bacteremia, underlying chronic diseases, and the severity of disease), microbiological data (monomicrobial or polymicrobial etiology, isolation of an enterococcal species, and susceptibility of the infecting organism to penicillin, ampicillin, imipenem, ciprofloxacin, vancomycin, teicoplanin, gentamicin, and streptomycin), clinical features (source of the bacteremia, severity of illness, and antimicrobial treatment used), and prognostic features (hospital stay after acquiring the enterococcal bacteremia and crude mortality).

Definitions. The isolation and identification of Enterococcus spp. were carried out by customary methods (5). Study of the susceptibilities of the organisms to different antimicrobial agents and the detection of HLRG and high-level resistance to streptomycin were performed according to the criteria of the National Committee for Clinical Laboratory Standards (21).

Statistical analysis. The level of resistance to gentamicin (the presence versus the absence of high-level resistance) was chosen as a dependent variable, and its possible association with the epidemiological, microbiological, clinical, and prognostic variables described previously was analyzed. In a first analysis, we studied the influence of the predisposing factors on the appearance of strains with HLRG (epidemiological and microbiological variables) for all strains collected. A univariate analysis was carried out by the chi-square test for qualitative variables (or two-tailed Fisher's exact test when some expected values were less than 5) and the Mann-Whitney U test for quantitative variables. Those variables for which an association with the appearance of HLRG was obtained were introduced in a multivariate analysis through a forward stepwise multiple logistic regression model, by means of the Wald test, for the determination of those variables that were presented as independent associations (11).

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Risk factors associated with HLRG. A total of 93 patients with bacteremia caused by Enterococcus spp. were included in the study; 31 of them were infected with strains with HLRG (33%) and 62 were infected with strains without HLRG (67%). Twenty-one patients were from community hospitals and 72 patients were from university hospitals, with no differences with respect to the rate of HLRG among patients from the two types of hospitals (P = 0.29).

View larger version (17K): [in this window] [in a new window]   FIG. 1.   Penicillin and ciprofloxacin MICs for strains with and without HLRG.

Influence of HLRG on prognosis. Eighty-eight patients without endocarditis were included in the crude mortality analysis. HLRG did not influence the outcome (9 of 31 patients with bacteremia caused by strains with HLRG [29%] versus 16 of 57 patients with bacteremia caused by strains without HLRG [28%]). An analysis of the probable influence of the poly or monomicrobial infection on mortality is presented in Table 4. It seems that different groups of factors are related to death in the groups of patients with mono- and polymicrobial infections, and previous ICU admission could not actually be related to death. Infection with a strain with HLRG was not a risk factor for death. The independent factors that were associated with excess of crude mortality were hemodynamic compromise, inappropriate antimicrobial therapy, and mechanical ventilation (Table 5).

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

The frequency of bacteremia caused by enterococci with HLRG is variable (7 to 63%), according to different investigators, and is related to the sources of specimens, the species considered, and the hospital (3, 8, 17, 23, 25, 34). Furthermore, the prevalence has been increasing in the last 15 years, as has been demonstrated in some long-term epidemiological follow-up studies (29, 34). In our series, one-third of the patients with enterococcal bacteremia were infected with strains with HLRG. This figure is similar to that found by other investigators in Spain in past years (4, 12, 15, 27, 33).

We have not found differences among patients with enterococcal bacteremias caused by strains with and without HLRG with respect to age, sex, or severity of the chronic underlying diseases. Disparity exists on this point in the literature, and while, as in our series, investigators have not found differences (23, 29), other investigators have isolated greater numbers of enterococcal strains with HLRG from those patients with more severe underlying diseases (34).

Chronic renal failure has been the only underlying disease or condition associated in the series with a greater frequency of bacteremias caused by strains with HLRG. This relationship has been described previously in relation to vancomycin-resistant strains (9, 32, 36).

Patients with bacteremia caused by strains with HLRG were grouped in the ICUs, while those in whom this characteristic was not found were more frequently grouped in medical wards. The results of other studies are disparate on this point, according to the different characteristics of the patients included in the study and the participating hospitals (3, 29). This makes a comparison between those studies and our series impossible.

Multiple studies support a relationship between the nosocomial acquisition and the development of enterococcal bacteremia caused by strains with HLRG (54 to 94% of all enterococcal bacteremias) (3, 8, 9, 23, 34, 35, 37). This relationship was also found in the present series and was independent of the duration of the previous hospital stay. However, according to the results of the multivariate analysis, it does not seem to be a risk factor by itself. It could be an expression of the fact that in this environment several factors present at the same time facilitate the appearance of strains with this characteristic (HLRG).

Those patients who had previously received antimicrobial agents had a five times greater risk of acquiring an enterococcal bacteremia caused by a strain with HLRG. Glycopeptides, quinolones, imipenem, and cephalosporins have been the antimicrobial agents related to the appearance of HLRG. This antecedent has been considered to be a risk factor or not to be a risk factor in different studies (3, 9, 23, 37). In a recent case-control study and by means of a multivariate analysis, the previous use of antimicrobial agents (and, more concretely, cephalosporins and aminoglycosides) was an independent risk factor for the acquisition of infections caused by -lactamase-producing, gentamicin-resistant E. faecalis strains (35).

Although one might think that previous ICU admission and previous administration of antimicrobial agents might be directly related, the multivariate analysis discards this possibility. Regarding previous ICU admission, the relationship found in our series did not depend on the previous duration of the stay in these units or the need for mechanical ventilation. Factors such as the presence of an intravascular catheter, urinary catheter, and nasogastric tube are not independent risk factors, and they may have been related to the grouping of the patients with enterococcal bacteremias caused by strains with HLRG in the ICUs.

Although we have not found any study in the literature that supports this fact, several investigators have demonstrated that ICUs have been the starting points of different epidemic outbreaks of infections caused by enterococcal strains with HLRG (alone or in association with resistance to other antimicrobial agents) (9, 33, 35).

Previous studies found that surgery was a risk factor for the acquisition of enterococcal bacteremias caused by strains with HLRG (35). We have not found a relationship between HLRG and a stay on surgical hospitalization wards or previous surgery. Again, local differences could explain this fact.

Although there was a higher proportion of severely ill patients among the patients with enterococcal bacteremias caused by strains with HLRG, this was not significant, and no other clinical feature differentiated the groups of patients. Other studies have identified an APACHE II index score of more than 6 points as a factor predicting infection with enterococcal strains with HLRG (35).

The frequencies of the sources of bacteremia in both groups were somewhat different, but they did not reach significance (bacteremias caused by strains with HLRG were more frequently of an unknown origin or were the result of the presence of an intravascular catheter). As in previous studies, we also found that endocarditis had a monomicrobial etiology and that the patients had community-acquired infections (13, 14, 25).

Contrary to other investigators who found a higher frequency of strains with HLRG among patients with polymicrobial bacteremia (34), we, among others, have not found such an association (3). The species most associated with HLRG was E. faecalis, and the total absence of E. faecium strains with HLRG is worth noting. There is a large range of HLRG according to the different enterococcal species (4 to 63% for E. faecalis and 2 to 24% for E. faecium) (3, 8, 17, 23, 33, 34), and the predominant species also differs. In some studies it is E. faecalis species (8, 33, 34), and in other studies it is E. faecium species (17). This special characteristic of our series makes E. faecalis species the most important independent risk factor for the development of HLRG. This is a very important epidemiological datum in our environment, although it may not be applicable to other geographical areas.

Most strains with HLRG have also been highly resistant to streptomycin. This is an interesting feature that is scarcely reported in Spain (4, 15) (rates from 7 to 59%, generally [27, 33]) but that is frequently found worldwide (41 to 82%) (3, 8, 17, 23, 26). It calls our attention to the possibility of a lack of alternative treatment for severe enterococcal infections in our environment. It is also important to emphasize that we have only found one strain resistant at the same time to ampicillin and to both aminoglycosides and that we have not found vancomycin-resistant strains. Enterococcal strains with HLRG were also less susceptible to penicillin and ciprofloxacin. This lower degree of susceptibility of enterococcal strains with HLRG to quinolones has been described previously (4, 22).

The crude mortality rate for the patients with enterococcal bacteremia without endocarditis has not been influenced by the susceptibility to gentamicin and is intermediate in comparison with that reported in the literature (from 18 to 61%) (24). This absence of an influence continued even after analyzing only the monomicrobial bacteremias (Table 4). This fact has also been recently reported by Patterson et al. (25), who found that only two factors were independent predictors of the outcomes for patients with enterococcal infection: resistance to ampicillin and the APACHE II index score. Factors which were determinants of the outcomes for patients in our series were hemodynamic compromise, inappropriate antimicrobial therapy, and the need for mechanical ventilation during the ICU stay. So, monotherapy with an active antimicrobial agent (penicillin or a glycopeptide) is probably sufficient for the treatment of enterococcal bacteremias without endocarditis.

The management of patients with enterococcal bacteremia without endocarditis has been discussed over the past 15 years in relationship to the clinical significance of such isolations. The mortality among these patients could be more related to the underlying conditions than to the enterococcal bacteremia (20). However, subsequent studies have outlined the need for an appropriate antimicrobial treatment for these patients to achieve a lower mortality rate (10, 13, 24, 25). The present study stresses this need and indicates that appropriate antimicrobial treatment is a very important factor in the patient's outcome. Patterson et al. (25) reported a mortality rate of 71% for patients who did not receive appropriate antimicrobial treatment, compared to a mortality rate of 53% for patients who did, although this difference did not reach statistical significance.

A polymicrobial etiology and the nosocomial acquisition of enterococcal bacteremia have been the independent risk factors for a long hospital stay after the acquisition of enterococcal bacteremia.