Summary of the unadjusted results of the 20 studies included in the meta-analysis. The estimated odds ratio (OR) and 95% confidence intervals (CIs) are shown for each study. The pooled OR is 4.5 (95% CI, 3.0-6.9). There was a significant heterogeneity across studies (P<.001).
Summary of the unadjusted results of the studies according to the criteria of selection of the control group. The estimated odds ratio (OR) and 95% confidence interval (CI) are shown for each study. The pooled OR for the 5 studies that used control patients from whom vancomycin-susceptible enterococci (VSE) had been isolated was substantially higher (pooled OR, 10.7; 95% CI, 4.8-23.8) than that of the group of 15 studies that used controls who had no vancomycin-resistant enterococci (VRE) isolated (pooled OR, 2.7; 95% CI, 2.0-3.8).
Summary of the unadjusted results of the studies according to the method of identifying case patients (excluding the studies that used controls with vancomycin-susceptible enterococci. The estimated odds ratio and 95% confidence interval are shown for each study. Pooled odds ratios were similar across the different methods.
Comparison of the results of studies that did not adjust for differences in length of hospitalization (LOS) between case and control patients and studies that did. The estimated odds ratio (OR) and 95% confidence interval (CI) are shown for each study. The studies that adjusted for the differences in LOS found only a small and nonsignificant association between vancomycin exposure and hospital-acquired vancomycin-resistant enterococci (OR, 1.4; 95% CI, 0.74-2.60).
Funnel plot used to assess publication bias. The left lower quadrant representing small studies with nonsignificant associations between vancomycin and hospital-acquired vancomycin-resistant enterococci is empty, suggesting publication bias.
Carmeli Y, Samore MH, Huskins WC. The Association Between Antecedent Vancomycin Treatment and Hospital-Acquired Vancomycin-Resistant EnterococciA Meta-analysis. Arch Intern Med. 1999;159(20):2461-2468. doi:10.1001/archinte.159.20.2461
Copyright 1999 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.1999
The association between vancomycin hydrochloride treatment and vancomycin-resistant enterococci (VRE) has been investigated in numerous studies with variable results.
To conduct a meta-analysis to estimate the magnitude of the association between vancomycin treatment and individual risk of VRE and to identify study characteristics that accounted for heterogeneity in study results.
Studies were identified using MEDLINE with index terms "Enterococcus," "Enterococcus faecalis," or "Enterococcus faecium" and "vancomycin," "drug resistance," "drug resistance, microbial," or "drug resistance, multiple or risk factors." Reports from conferences and reference lists of recent reviews were used. A total of 420 published reports and 98 conference reports were reviewed; 20 studies described in 15 published reports were included in the analysis. We recorded study period, hospital setting, case and control definitions, length of hospital stay, method of adjustment for differences in length of stay, and data on treatment with vancomycin. The odds ratio (OR) of vancomycin treatment provided the measure of association analyzed. A random-effects model was used to estimate the pooled OR.
When results from all 20 studies were combined, the pooled OR was 4.5 (95% confidence interval, 3.0-6.9), but the test for heterogeneity was highly significant (P<.001). The 5 studies that used patients with vancomycin-susceptible enterococci as controls found a stronger association (pooled OR, 10.7; 95% confidence interval, 4.8-23.8) than the 15 studies that used controls who had no VRE isolated (pooled OR, 2.7; 95% confidence interval, 2.0-3.8). After restricting the analysis to the latter studies only, no heterogeneity was evident in the unadjusted study results. Patients with VRE had stayed in the hospital much longer than control patients. Studies that adjusted for this difference found only a small and nonsignificant association between vancomycin treatment and VRE (pooled OR, 1.4; 95% confidence interval, 0.74-2.60). We also detected publication bias, favoring report of studies that found a large measure of association.
The reported strong association between vancomycin treatment and hospital-acquired VRE results from the selection of the reference group, confounding by duration of hospitalization, and publication bias. Studies that accounted for these factors found only a small and nonsignificant association.
FIRST ISOLATED in 1987,1,2 vancomycin hydrochloride–resistant enterococci (VRE) have rapidly become established as important nosocomial pathogens in the United States. In some hospitals, VRE are responsible for greater than 20% of enterococcal infections.
Initial epidemiological studies3,4 identified therapy with vancomycin as a risk factor for VRE infection or colonization. The emergence of VRE was also coincident with dramatic increases in global vancomycin use in the 1980s and early 1990s. For example, investigators5 in a large university hospital documented a 20-fold increase in the use of intravenous vancomycin during the 10-year period from 1981 to 1991.
These data were widely interpreted as indicating that vancomycin treatment facilitated acquisition of VRE by individual patients and that increased use of vancomycin played a major role in the emergence and dissemination of VRE in the United States. This conclusion was consistent with the broader concept that antibiotic treatment exerts selection pressure that promotes resistance. Consequently, interventions to facilitate "prudent vancomycin use" were emphasized strongly in recommendations for preventing the spread of VRE published by the Hospital Infection Control Practices Advisory Committee of the Centers for Disease Control and Prevention in 1995.6
However, subsequent studies of hospital-acquired VRE reported variable results regarding the association between vancomycin treatment and VRE. Many of the investigations were of relatively small size and so of limited power to detect statistically significant effects of vancomycin treatment. In addition, some studies have demonstrated an association between VRE and antibiotic agents other than vancomycin, such as cephalosporins or metronidazole.
Our major objectives in this meta-analysis were to more precisely estimate the magnitude of the association between vancomycin treatment and individual risk of VRE and to identify study characteristics that accounted for heterogeneity in results; specifically, we examined the methods used for identification of case patients and the criteria for selection of control (comparison) patients and determined whether there was an adjustment for length of hospital stay. We regarded length of stay as a particularly important confounding variable because it represents the duration of the at-risk period for exposure to vancomycin and for acquisition of VRE and is also a correlate of severity of illness. We also examined the role of publication bias (potential lack of reporting of small studies with nonsignificant results).
We defined a report as a printed document by an organization or group of authors. We defined a study as an analysis of data from case and control (comparison) patients. An individual report could describe more than 1 study.
We identified reports published in journals using a computerized literature search (MEDLINE). We used medical subject headings in the following search strategy: "Enterococcus," "Enterococcus faecalis," or "Enterococcus faecium" and "vancomycin," "drug resistance," "drug resistance, microbial," or "drug resistance, multiple or risk factors." We limited the search to reports on human subjects with an abstract in English published between January 1987 and March 1998. This search identified 420 published reports.
We identified reports from conferences by searching the official conference proceedings of the 1996 and 1997 annual meetings of the Infectious Diseases Society of America, the Interscience Conference on Antimicrobial Agents and Chemotherapy, and the Society for Healthcare Epidemiology of America. This search identified 98 conference reports.
We compared the list of reports identified by these searches with the reference lists of recent reviews.6- 13 No additional reports were identified.
We defined criteria for the inclusion and exclusion of studies before reviewing specific reports.
We included studies that (1) provided data regarding antecedent vancomycin therapy among patients who were infected or colonized with VRE and a concurrent control group of patients selected from the same general population at risk (eg, same hospital or unit or same primary diagnosis) and (2) were performed in hospitals in the United States. The latter criterion was based on the observation that the epidemiological features of VRE in Europe have been distinctly different from those in the United States regarding spread in the community compared with hospitals.
Since our intent was to examine the association of antecedent vancomycin therapy with infection or colonization with VRE at the level of the patient, we excluded studies examining the correlation between global vancomycin use and rates of infection or colonization with VRE. We also excluded studies that focused on enterococcal strains in which vancomycin resistance was shown to be conferred by the VANC gene or Enterococcus species in which vancomycin resistance is generally due to the VANC gene (Enterococcus casseliflavus, Enterococcus gallinarum, and Enterococcus flavescens) because the epidemiological characteristics of infection or colonization with these microorganisms differ substantially from those in which resistance is conferred by the VANA or VANB genes.13,14
Two of us (Y.C. and W.C.H.) independently reviewed the abstracts of all reports identified by the searches described above. Abstracts describing studies performed in hospitals outside the United States and studies focused solely on the results of antimicrobial susceptibility testing and/or molecular mechanisms of resistance among enterococci were not reviewed further. The entire text of the remaining reports was independently reviewed by both investigators who recorded predetermined information relevant to the inclusion and exclusion criteria. The investigators compared their evaluations, discussing discrepancies and reviewing studies again until they reached a consensus regarding satisfaction of these criteria.
Two of us (Y.C. and W.C.H.) extracted the following information regarding each study, independently using a data extraction form: year of publication, period of study, hospital unit or primary diagnosis of the study population, method used to identify case patients, method used to select control patients, method used to adjust for length of hospital stay, the total number of case and control patients and the number who were exposed to vancomycin, the odds ratio (OR) and confidence intervals (CIs) (crude and adjusted, if reported) for antecedent vancomycin therapy, the average number of days in the hospital (before a culture for VRE was obtained) for case patients, and the day of inclusion for controls. In the few reports in which the number of patients treated with oral vancomycin was reported separately from the number of patients treated with intravenous vancomycin, only the number of patients treated with intravenous vancomycin was used in the analysis. The investigators compared their evaluations, discussed discrepancies, and reviewed studies again together as necessary. Authors of the original reports were contacted for clarification when incomplete or ambiguous data were reported.
We classified the method for identifying case patients into 1 of 3 categories: (1) clinical cultures—case patients detected through clinical cultures ordered by clinicians as part of patient care, (2) surveillance—case (and control) patients identified through systematic surveillance cultures (cross-sectional identification), and (3) acquisition—case patients who underwent serial culturing and had a negative VRE culture result followed by a positive culture result (incident cases).
We classified the method for selection of control patients as either "no VRE" or "vancomycin-susceptible enterococci (VSE)." No VRE was defined as control patients who were not identified as colonized or infected with VRE by the methods described above. Control patients in these studies typically represented a sample of the hospital or unit population. Vancomycin-susceptible enterococci was defined as control patients who were infected or colonized with vancomycin-susceptible strains of enterococci.
We classified the method of adjustment for length of hospital stay as either "adjustment" or "no adjustment." Adjustment was defined as either matching of case and control patients regarding length of hospital stay or use of multivariable analysis to adjust for differences in length of hospital stay. No adjustment was defined as the use of neither of these 2 methods.
We managed reports using computer software (EndNote, version 3.0; Niles Software, Inc, Berkeley, Calif). We calculated the crude OR and 95% CIs by the Woolf approximation using the primary data reported for the study. One case was added to empty cells to allow calculation of the OR. If the primary data were not reported, we used the OR and 95% CIs reported for the study.
We used the DerSimonian and Laird15 random-effect model to obtain pooled estimates of the OR and 95% confidence limits for a group of studies. We tested for heterogeneity in the results of different studies using the Q statistic and considered heterogeneity to be significant if P<.25.
Publication bias was evaluated by plotting the point estimate against the percentage weight of the study (1/variance, a measure of the power of the study) and constructing a funnel diagram originating from the pooled OR of the studies that weighted more than the average study weight, and included as many studies as possible.
The Egger et al16 regression asymmetry test and the Begg and Mazumar17 adjusted rank correlation test were used to test for publication bias. Paired t tests were used to test the differences in length of hospitalization between case and control patients.
Twenty case-control analyses examining the association of vancomycin treatment and VRE were described in 15 published reports that met the inclusion criteria.18- 32 One report29 described 2 studies using part of the same case group but different control groups. For the purpose of this analysis, we considered these 2 studies to be independent. Characteristics of the studies are outlined in Table 1 and discussed further. Twenty-three other relevant publications not meeting the study criteria and the main reason for noninclusion are listed in the "Suggested Reading" section.
All of the studies included were designed to examine simultaneously several possible risk factors for VRE, such as location in the hospital, length of hospital stay, comorbidities, and exposure to antibiotics, including vancomycin. Vancomycin exposure was predominantly treatment with intravenous vancomycin. Studies were performed between 1990 and 1995 and included various patient populations; 8 were hospital-wide studies, 7 were intensive care unit studies, and 3 were oncology studies. Nine of the studies identified case patients using results of clinical cultures; blood was the source of the clinical culture in 4 of these studies. Seven studies identified case patients by surveillance. Three studies identified patients who acquired VRE using serial surveillance cultures. Five studies used controls with VSE; all of these studies identified case patients using clinical culture results.
Pooling the raw data from every study, a total of 586 patients with hospital-acquired VRE, of whom 61% were treated with vancomycin, were compared with 1249 controls, of whom 26% were treated with vancomycin. All of the reports described an unadjusted positive association of exposure to vancomycin and colonization or infection with VRE; odds ratios ranged from 1.3 to 44.0 (Figure 1). The pooled OR for the unadjusted results of these studies was 4.5 (95% CI, 3.0-6.9). Ten studies found an unadjusted significant association, and 10 others did not find a significant association between vancomycin treatment and VRE. There was a significant heterogeneity among studies' results (P<.001). We studied potential explanations for heterogeneity in the results.
Two different strategies were used for the selection of controls. Five studies used controls who had vancomycin-sensitive enterococci isolated from their cultures. Fifteen other studies used controls who had no VRE isolated from their cultures. The pooled OR for the first group of studies was substantially higher (pooled OR, 10.7; 95% CI, 4.8-23.8) than that of the group of 15 studies that used controls who had no VRE isolated (pooled OR, 2.7; 95% CI, 2.0-3.8) (Figure 2).
We also examined the hypothesis that the association for exposure to vancomycin would vary with case definition. Studies were divided into 3 groups based on case definitions: clinical cultures, surveillance, and acquisition. The method of case identification appeared to account for part of the heterogeneity in results; studies identifying case patients based on clinical culture results found a stronger association between vancomycin and VRE (pooled OR, 6.2; 95% CI, 3.2-12.2) than studies identifying case patients by surveillance (pooled OR, 3.3; 95% CI, 1.7-6.3) or acquisition (pooled OR, 3.2; 95% CI, 1.7-5.9). There was significant heterogeneity in the results of the first 2 groups of studies (P<.001 and P = .08, respectively), whereas the results of studies examining acquisition were homogeneous (P = .98). On further analysis, to explain heterogeneity, the high pooled OR for vancomycin in studies that relied on clinical culture results for case detection was due to the association between use of controls with VSE and use of case patients detected by clinical culture results. When the studies using controls with VSE were excluded, this method of case identification gave similar results to the other methods (pooled OR, 2.0; 95% CI, 1.2-3.5), and no heterogeneity was found between the combined unadjusted results (Figure 3).
Since the study design of using controls with VSE was the major source of heterogeneity and appeared to be associated with distorted effects, such studies were excluded from further analyses.
In the remaining group of 15 studies, we compared the strength of the association between exposure to vancomycin and VRE in 2 groups of studies: 10 studies that did not adjust for length of hospitalization and 5 studies that did adjust for length of hospitalization by design or by analysis (Figure 4). The pooled OR for studies that did not adjust for length of hospitalization was 3.1 (95% CI, 1.8-5.3); for studies that did adjust for length of hospitalization, 1.4 (95% CI, 0.74-2.60). The results of the studies that adjusted for length of stay were not heterogeneous (P = .65).
Seven of the 10 studies that did not adjust for the length of stay nevertheless reported specific data on duration of hospital stay for case and control patients. In these studies, a substantial difference in the duration of hospital stay between case and control patients was found. Case patients were hospitalized until study inclusion a mean ± SD of 16.0 ± 5.8 days longer than controls (P = .03).
We examined the relation between the weight of the studies (a combination of the study size and the number of outcomes, a measure of the power of the study) and the unadjusted OR reported by the study. Low-weight studies (low power) were more likely to report high ORs and significant results than large studies (high power). The pooled OR was 4.4 (95% CI, 2.7-7.1) for the 11 studies with less than 7% weight (the average study weight), whereas the pooled OR was 1.8 (95% CI, 1.2-2.7) for the 4 studies with a weight greater than 7%. Figure 5 demonstrates that small studies with modest effects appear missing from the left lower quadrant of the funnel graph. The regression asymmetry test and the adjusted rank correlation test were compatible with the presence of publication bias (P<.01 and P = .13, respectively).
The rationale for conducting this meta-analysis of the association between antecedent vancomycin treatment and hospital-acquired VRE was that results of previous studies have been inconsistent and that many have had relatively small sample sizes. The subject of nosocomial VRE and its relation to vancomycin is one of intense interest in the medical community, as evidenced by the large number of studies that have been conducted during a relatively short time. The postulated causal relation between vancomycin treatment and VRE has largely motivated efforts to devise effective means of reducing vancomycin use.6,33,34
The main findings of this meta-analysis were the following: (1) although the crude analysis of all published studies appeared to provide strong support for the association between vancomycin and VRE, there was significant heterogeneity among studies results; (2) the major sources of heterogeneity were related to study design regarding selection of controls and lack of adjustment for confounding by length of stay; (3) once confounding and other sources of heterogeneity were taken into account, the association between vancomycin and VRE was small and not statistically significant; and (4) a lack of small studies with modest effect sizes was discerned, suggesting the presence of some degree of publication bias. Studies of low power with nonsignificant results may have been less likely to be published, or if published, less likely to report in detail the association between vancomycin and VRE.
Studies that used controls with VSE showed a substantially stronger association for antecedent vancomycin treatment compared with other studies (pooled OR, 10.7 vs 2.7). We believe that a study design that uses patients with VSE as controls has inherent shortcomings regarding identification of causal risk factors. Control patients with VSE likely represent a biased sample of the source population of hospitalized patients, at least regarding exposures that may influence the development of VSE. Treatment with vancomycin probably reduces the likelihood of subsequent infection with VSE, thus lowering the frequency of antecedent vancomycin exposure in these patients relative to the general hospital population and inflating the magnitude of vancomycin effect in studies that used this group as controls.
The other major factor that gave rise to an exaggerated association between vancomycin and VRE was confounding by length of hospital stay. Five studies that adjusted for length of hospital stay exhibited a pooled OR of 1.4, with the lower bound of the 95% CI extending to 0.74. In studies that did not adjust for length of stay but nevertheless provided detailed data, case patients with VRE had substantially longer lengths of stay than controls. This is expected based on the relation between days in the hospital and cumulative risk of acquisition of a nosocomial pathogen such as VRE. Conversely, in contrast to our a priori hypothesis, we did not observe heterogeneity in ORs according to whether case patients were identified by clinical culture, surveillance, or acquisition.
One implication of this meta-analysis is that if vancomycin is not a significant causal factor in nosocomial VRE, curtailment of vancomycin use may have relatively little impact on incidence rates of VRE infection or colonization. Other antibiotics, such as cephalosporins or antianaerobic agents, may play an equally if not more important role in promoting the spread of VRE within institutions. Given the complex genetic machinery required to confer vancomycin resistance, de novo emergence of resistance is unlikely in an individual patient.35 Thus, newly detected VRE may represent either acquisition of resistant organisms or expansion of preexisting but undetected populations of VRE.36 Therefore, antibiotics may facilitate detection of VRE largely through effects on competing gastrointestinal flora, for which the action of vancomycin may be less prominent than other antibiotic agents.
We focused our analysis on only a few study characteristics. Other characteristics, such as the methods used to determine the susceptibility of enterococcal strains to vancomycin or variation in the source patient population, were not explored. The homogeneity found in the unadjusted study results, after excluding the studies that used controls with VSE, suggests that we accounted for the most important areas of heterogeneity. Most important, results of studies did not vary according to case definition, despite the different microbiologic methods used in these different types of studies.
Intravenous vancomycin constituted the vast majority of vancomycin use and, therefore, insufficient data existed to specifically examine the effect of oral vancomycin. Similarly, lack of available data prevented an examination of the effect of duration of vancomycin exposure. Furthermore, an analysis of the effects of antibiotics other than vancomycin was precluded by the limited amount of data and was beyond the scope of this study.
The results of this analysis do not abrogate the notion that vancomycin played a crucial role in the initial emergence of vancomycin resistance. Moreover, antibiotic selection pressure may have different effects during various stages of an epidemic. We also acknowledge that this meta-analysis was limited to studies conducted in the United States and, therefore, the findings may not be generalizable to countries where patterns of spread have been distinctly different. Finally, vancomycin treatment may have indirect effects on the transmission and acquisition of VRE that might be detectable only in ecological, population-level studies.
In summary, our results suggest that the reported association between antecedent vancomycin treatment and VRE is distorted by the method of selection of the control group, lack of adjustment for duration of hospitalization, and publication bias. After taking these factors into account, the pooled association between vancomycin and VRE is of modest size and not statistically significant. Thus, we conclude that, at most, vancomycin has a small degree of association with hospital-acquired VRE when examined at the level of the individual patient.
Anglim AM, Klym B, Byers KE, Schled WM, Farr BM. Effect of a vancomycin restriction policy on ordering practices during an outbreak of vancomycin-resistant Enterococcus faecium. Arch Intern Med. 1997;157:1132-1136. [No data on vancomycin exposure (data from the same study published in an abstract that is included in the meta-analysis).]
Bonten MJM, Hayden MK, Nathan C, Rice TW, Weinstein RA. Stability of vancomycin-resistant enterococcal genotypes isolated from long-term–colonized patients. J Infect Dis. 1998;177:378-382. [Not a risk-factors analysis.]
Breezhold DW, Slaughter S, Hayden MK, et al. Skin colonization with vancomycin-resistant enterococci among hospitalized patients with bacteremia. Clin Infect Dis. 1997;24:704-706. [No exact data on vancomycin exposure.]
Chadwick PR, Oppenheim BA, Fox A, Woodford N, Morgenstern GR, Scarffe JH. Epidemiology of an outbreak due to glycopeptide-resistant Enterococcus faecium on a leukemia unit. J Hosp Infect. 1996;34:171-182. [Not a US study. No control group.]
Edmond MB, Ober JF, Dawson JD, Weinbaum DL, Wenzel RP. Vancomycin-resistant enterococcal bacteremia: natural history and attributable mortality. Clin Infect Dis. 1996;23:1234-1239. [No data on vancomycin exposure.]
Frieden TR, Munsiff SS, Low DE, et al. Emergence of vancomycin-resistant enterococci in New York City. Lancet. 1993;342:76-79. [No control group.]
Gordts B, Van Landuyt H, Ieven M, Vandamme P, Goossens H. Vancomycin-resistant enterococci colonizing the intestinal tract of hospitalized patients. J Clin Microbiol. 1995;33:2842-2846. [Not a US study.]
Green M, Barbadora K, Michaels M. Recovery of vancomycin-resistant gram-positive cocci from pediatric liver transplant recipients. J Clin Microbiol. 1991;29:2503-2506. [Mostly VANC resistance and nonenterococcal organisms.]
Green M, Wadowsky RM, Barbadora K. Recovery of vancomycin-resistant gram-positive cocci from children. J Clin Microbiol. 1990;28:484-488. [Most of the organisms in the study were not enterococci.]
Karanfill LV, Murphy M, Josephson A, et al. A cluster of vancomycin-resistant Enterococcus faecium in an intensive care unit. Infect Control Hosp Epidemiol. 1992;13:195-200. [No concurrent control group.]
Lam S, Singer C, Tucci V, Morthland VH, Pfaller MA, Isenberg HD. The challenge of vancomycin-resistant enterococci: a clinical and epidemiologic study. Am J Infect Control. 1995;23:170-180. [No control group.]
Livornese LL, Dias S, Samel C, et al. Hospital-acquired infection with vancomycin-resistant Enterococcus faecium transmitted by electronic thermometers. Ann Intern Med. 1992;117:112-116.[No exact data on vancomycin exposure.]
Luber AD, Jacobs RA, Jordan M, Guglielmo BJ. Relative importance of oral versus intravenous vancomycin exposure in the development of vancomycin-resistant enterococci. J Infect Dis. 1996;173:1292-1293. [No individual data. No control group.]
Montecavalo MA, Horowitz H, Gedris C, et al. Outbreak of vancomycin-, ampicillin-, and aminoglycoside-resistant Enterococcus faecium bacteremia in an adult oncology unit. Antimicrob Agents Chemother. 1994;38:1363-1367. [No appropriate control group (comparing VRE bacteremia with VRE stool carriage).]
Montecavalo MA, deLencastre H, Carraher M, et al. Natural history of colonization with vancomycin-resistant Enterococcus faecium. Infect Control Hosp Epidemiol. 1995;16:680-685. [No data on vancomycin exposure. No control group.]
Montecavalo MA, Shay DK, Patel P, et al. Bloodstream infections with vancomycin-resistant enterococci. Arch Intern Med. 1996;156:1458-1461. [No exact data on number of patients treated with vancomycin.]
Moreno F, Grota P, Crisp C, et al. Clinical and molecular epidemiology of vancomycin-resistant Enterococcus faecium during its emergence in a city in southern Texas. Clin Infect Dis. 1995;21:1234-1237. [No exact data on vancomycin exposure.]
Papanicolaou GA, Meyers BR, Meyers J, et al. Nosocomial infections with vancomycin-resistant Enterococcus faecium in liver transplant recipient: risk factors for acquisition and mortality. Clin Infect Dis. 1996;23:760-766. [Vancomycin exposure selectively reported.]
Quale J, Landman D, Atwood E, et al. Experience with a hospital-wide outbreak of vancomycin-resistant enterococci. Am J Infect Control. 1996;24:372-379. [No control group.]
Quale J, Landman D, Saurina G, Atwood E, Ditore V, Patel K. Manipulation of hospital antimicrobial formulary to control an outbreak of vancomycin-resistant enterococci. Clin Infect Dis. 1996;23:1020-1025. [No individual data. No concurrent control group.]
Wade JJ. The emergence of Enterococcus faecium resistant to glycopeptides and other standard agents: a preliminary report. J Hosp Infect. 1995;30(suppl):483-493. [Not a US study. No exact data on vancomycin use.]
Weinstein JW, Roe M, Towns M, et al. Resistant enterococci: a prospective study of prevalence incidence, and factors associated with colonization in a university hospital. Infect Control Hosp Epidemiol. 1996;17:36-41. [The outcome examined was a mixture of VRE and ampicillin-resistant enterococci.]
Wells CL, Juni BA, Cameron SB, et al. Stool carriage, clinical isolation, and mortality during an outbreak of vancomycin resistant enterococci in hospitalized medical and/or surgical patients. Clin Infect Dis. 1995;21:45-50. [No control group. No vancomycin data.]
Accepted for publication March 2, 1999.
Reprints: Yehuda Carmeli, MD, MPH, Division of Infectious Diseases, Tel Aviv Sourasky Medical Center, 6 Weizman St, Tel Aviv 64239, Israel (e-mail: firstname.lastname@example.org).