Using the rate of bacterial clearance determined by real-time polymerase chain reaction to evaluate antibiotic effectiveness in Acinetobacter baumannii bacteremia.

B. Decker,J. Powers

DOI: https://doi.org/10.1097/CCM.0b013e318283c34a

IF: 8.8

2013-05-01

Critical Care Medicine

Abstract:To the Editor: Chuang et al (1) report the use of quantitative real-time polymerase chain reaction (qRT-PCR) of Oxa-51 DNA in patients with Acinetobacter baumannii bacteremia. Comparing qPCR results of patients who received antibiotics with in vitro activity to those given antibiotics with less activity, they noted an association between increased rate of clearance of blood bacterial DNA and antibiotic choice. The authors concluded that qPCR rate of bacterial clearance is a useful “surrogate marker for evaluating different antimicrobial regimens in future studies” (1). Data suggesting associations between the biomarker Oxa51 DNA blood clearance and appropriate antibiotic use do not necessarily demonstrate a causal relationship between clearance and antibiotic effectiveness. As defined by the Biomarkers Definitions Working Group (2), biomarkers are clinical tools that can be used in different ways for diagnosis, prognosis, or replacement endpoints for direct outcomes like mortality. Biomarkers used for prognosis can be evaluated in this type of retrospective observational study. However, from this study, we cannot make the assumption that improvement in clearance rate was due to the antibiotics chosen. This association may be due to baseline patient differences independent of treatment. These differences ultimately can be more important in determining outcomes. Patients given active antibiotics received them because they had less resistant organisms. Patients with less resistant organisms are different than patients infected with susceptible organisms in ways that affect outcomes. The reported relationship is therefore confounded. Prentice (3) has noted that in order to represent a true endpoint (in this case ICU mortality) the treatment difference on the surrogate biomarker in a baseline comparable group should reflect the treatment difference on the true endpoint. Baseline comparability is usually achieved through randomization. It often is not possible to account for measured and unmeasured differences in patient factors in observational databases. Finally, it is not clear mechanistically that persistent bacteremia is directly the cause of death in these patients. Munford (4) has discussed this uncertainty. Bacteremia, or DNA detectable by qPCR, may be an indicator of uncontrolled pneumonia, as this was the most common source of Acinetobacter infection. In this case, pneumonia is the cause of bacteremia and death. Without assurance of a causal relationship between qPCR results and mortality, we risk using a valid prognostic tool, such as qPCR rate of change (patients who do not clear bacteria have worse outcomes) as an invalid measure of antibiotic effectiveness (because one antibiotic increases qPCR rate of clearance, it will lower mortality more than another antibiotic). Overextrapolating this association into a causal relationship can lead to poor decision making. We appreciate the use of Oxa-51 qPCR as a potentially rapid test of A. baumannii bloodstream infection. The suggested association with prognosis may be useful in clinical practice. Due to the significant risk of confounding, it would be premature to use qPCR of Oxa-51 as a surrogate endpoint to evaluate effectiveness of antibiotic therapy until this biomarker is properly evaluated to reflect effects on mortality in randomized trials. Supported, in whole or in part, with federal funds from the National Cancer Institute, National Institutes of Health, under contract HHSN261200800001E. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government. The authors have not disclosed any potential conflicts of interest.