Criteria abstracted from The Users' Guide to Medical Literature, from the Health Information Research Unit and Clinical Epidemiology and Biostatistics, McMaster University

Highlighted lines and questions below provide links to the pertinent description of criteria in The EBM User's Guide, now available at the Canadian Centres for Health Evidence

Article Reviewed:

Please visit the new Evidence Based Journal Club Reviews

Does the empiric use of vancomycin in pediatrics increase the risk for Gram-negative bacteremia?

Van Houten MA, Uiterwaal CS, Heesen GJ, Arends JP, Kimpen JL.

Pediatr Infect Dis J 2001;20(2):171-7. [abstract]

 

Reviewed by Curtis Kennedy, MD , Baylor College of Medicine, Pediatric Critical Care Section, Texas Children's Hospital, Houston, TX

Review posted August 13, 2002

I. What is being studied?:

The study objective:

To quantify the risk of Gram-negative bacteremia associated with the empiric use of vancomycin and other classes of antibiotics.

The study design:

Retrospective case control study

The patients investigated:

Hospitalized patients treated between 1/1/1994 and 12/31/1997 (4 years) in whom a blood culture was performed because of suspected bacteremia. Subjects were all (N=105) patients with a blood culture that grew a Gram-negative organism. Controls were a random (N=225) subset of patients with a negative blood culture.

The patients excluded:

3 positive blood cultures (of 108 specimens) were excluded because the specimens were from patients with previous cultures. 4 negative blood cultures (of 229 specimens) were excluded for similar reasons.

The treatment, intervention, or exposure in question: :

Vancomycin; other classes of antibiotics, including: aminoglycosides, penicillins and cephalosporins; TISS 28 score (indicator for intensity of care); steroids (dexamethasone); surgical status; central venous catheter use; parenteral nutrition; mechanical ventilation; and other various variables. (Summarized in Table 2).

II. Are the results of the study valid?

Primary questions:

1.Were there clearly identified comparison groups that were similar with respect to important determinants of outcome, other than the one of interest?

A) The control population did not well match the case population: the chief problem was the difference in TISS scores (13 for controls vs. 19 for cases). B) In the results section, the authors allude to a difference in age, with controls being younger. This is not supported by Table 1. C) Also, if one assumes that the primary diagnosis does not change during the hospitalization, then the case population contained a greater percentage of neonates and the control population contained a greater percentage of cardiology, pulmonology, congenital and neurology cases.

Of note, the authors indicate concern that the case population had a longer duration of admission. This value is listed in the clinical follow-up section of Table 1, and therefore we feel their concern regarding this fact is unwarranted: patients with Gram-negative bacteremia are expected to be more ill and require longer hospitalization than those without. It does not necessarily impact the validity of the study.

We feel this study would have had greater validity if the control group been matched to the exposure group (with respect to age, diagnosis and TISS score). To accomplish this, the authors would need to specify age, diagnosis and TISS as criteria variables and take a random sample of negative controls for comparison rather than selecting a random sample from all negative blood cultures.

2. Were the exposures and outcomes measured in the same way in the groups being compared?

We feel this is the major strength of the study. The authors did a good job of controlling bias by giving a priori definitions of variable assignment: 10 days preceding culture, any exposure, yes/no exposure classification. Although dose dependent changes are lost by this method, we would not have changed the study design.

3. Was follow-up sufficiently long and complete?

The outcome of interest was defined as inclusion criteria. (This is a case control study, not a cohort study.) Follow-up was therefore sufficient to evaluate the outcome.

Secondary questions:

4. Is the temporal relationship correct?

With respect to choosing 10 days prior to the event as a cutoff criteria, we feel this was an appropriate timeframe. However, we give caution to the wording on P. 175, first paragraph, next to last sentence: "We used only those risk indicators that were noted in the patients' charts before the physician's knowledge of the results of the blood culture." By this wording, it would be possible to have no prior exposure to vancomycin, have a blood culture taken for suspected infection, be placed on empiric vancomycin coverage, and obtain a "positive" result 24 to 48 hours later. Based on the above wording, vancomycin exposure would be classified as "yes." The measurements paragraph of the methods section identifies the start cutoff ("... maximally 10 days before the venipuncture for blood culture was taken") but does not explicitly state the "end" cutoff. Presumably, they only used information up to the point of the venipuncture, but that caveat is not explicitly stated in the methods section.

5. Is there a dose-response gradient?

This question is not applicable to this article. The variable in question (vancomycin) was dichotomous, thereby excluding it from this analysis.

III. What are the results?

1. How strong is the association between exposure and outcome?

Odds ratio for vancomycin was significant: OR=8.06 [95% CI = 3.1 to 20.9]. So, patients that received vancomycin were 8 times as likely to develop Gram-negative bacteremia as those that did not receive vancomycin. These findings were statistically significant.

The same held true even during bivariate and multivariate analysis. Odds ratios for vancomycin controlling for a single other variable (gender, antibiotic, steroid, surgery, etc) ranged from 5.94 to 8.5 [95% CI = 2.22 to 22.24] Odds ratios for vancomycin controlling for multiple other variables (medications, non-pharmacologic variables, and all variables combined) ranged from 3.88 to 6.7 [95% CI = 1.34 to 17.75]. Again, the findings were statistically significant.

2. How precise is the estimate of the risk?

The 95% CI ranged from 3.1 to 20.9. While this is not a very precise number (i.e., wide range), it is still significant. So, with 95% confidence we can say that the real risk of developing Gram-negative bacteremia is between 3 and 21 times greater if there was exposure to vancomycin.

IV. What are the implications for my practice?

1. Are the results applicable to my practice?

The patient population in this study shares many features of our own patient population: a mix of patients with no previous hospital days as well as those already hospitalized. Likewise, there was a broad representation of primary pathologies. However, the patient population did differ from our own with respect to the high percentage of neonates.

2. What is the magnitude of the risk?

The odds of a gram negative infection are 8 times higher in patients receiving vancomycin versus those who are not. We are unable to determine the absolute magnitude of the risk since we do not have the total number of blood cultures taken. (This is due to this being a case control study, not a cohort study.)

3. Should I attempt to stop the exposure?

Issue 1: Empiric nosocomial coverage: Given that as high as 60% of PICU blood stream infections are Gram-positive in nature (1) and given that as much as 40% of Staph. aureus species are methicillin resistant (2), we feel the potential harms associated with vancomycin exposure are outweighed by the potential benefits of adequate Gram-positive (MRSA) coverage. Therefore, it is our opinion that there is insufficient evidence to change our practice of empiric nosocomial antibiotic coverage with respect to vancomycin. However, it may be prudent to minimize the use of empiric vancomycin unless MRSA is suspected.

Issue 2: Empiric community-acquired coverage: Community Onset Staph Aureus (COSA) infections are fairly rare (incidence approximately 17 cases / 100,000 population) but are much more common in patients with recent health care exposure or chronic medical conditions. (3). Serious consideration to not starting empiric vancomycin coverage for patients presenting from the community who are not at high risk and do not appear toxic.

References

1. Gray J, Gossain S, Morris K. Three-year survey of bacteremia and fungemia in a pediatric intensive care unit. Pediatr Infect Dis J. 2001 Apr;20(4):416-21. [abstract]
2. Rezende NA, Blumberg HM, Metzger BS, et al. Risk factors for methicillin-resistance among patients with Staphylococcus aureus bacteremia at the time of hospital admission. Am J Med Sci. 2002 Mar;323(3):117-23. [abstract]
3. Morin CA, Hadler JL. Population-based incidence and characteristics of community-onset Staphylococcus aureus infections with bacteremia in 4 metropolitan Connecticut areas, 1998. J Infect Dis. 2001 Oct 15;184(8):1029-34. [abstract]

Comments

Back to the EB Journal Club Index