The trusted source for
healthcare information and
Can Procalcitonin Distinguish Sepsis from other Causes of SIRS?
Abstract & Commentary
By Andrew M. Luks, MD, Pulmonary and Critical Care Medicine, University of Washington, Seattle. Dr. Luks reports no financial relationship to this field of study.
This article originally appeared in the June 2007 issue of Critical Care Alert. It was edited by David J. Pierson, MD, and peer reviewed by William Thompson, MD. Dr. Pierson is Professor, Pulmonary and Critical Care Medicine, Harborview Medical Center, University of Washington, and Dr. Thompson is Staff Pulmonologist, VA Medical Center; Associate Professor of Medicine, University of Washington. Drs. Pierson and Thompson both report no financial relationships relevant to this field of study.
Synopsis: A meta-analysis of 18 studies that have examined the usefulness of procalcitonin measurement in the diagnosis of sepsis finds that the diagnostic performance of this measurement is low and that the test cannot reliably distinguish sepsis from other causes of the systemic inflammatory response syndrome.
Source: Tang BMP, et al. Accuracy of procalcitonin for sepsis diagnosis in critically ill patients: Systemic review and meta-analysis. Lancet Infect Dis. 2007;7(3):210-217.
Several studies have suggested that procalcitonin may be a useful biochemical marker to differentiate sepsis from other non-infectious causes of the systemic inflammatory response syndrome (SIRS), such as pancreatitis. Tang and colleagues conducted a meta-analysis of these studies in order to evaluate the diagnostic accuracy of procalcitonin in the diagnosis of sepsis, as well as to determine whether the increasing reliance on the test in clinical practice is, in fact, justified.
Out of a total of 39 candidate studies, 18 investigations encompassing 2097 patients were included in the final meta-analysis. Studies were excluded if they lacked a well-defined gold standard for the diagnosis of sepsis, provided insufficient information to create 2 x 2 contingency tables, and examined patients without SIRS or critical illness, or examined too narrow a spectrum of critically ill patients. The 18 studies were divided into 2 groups: Phase II studies, which examined how the test discriminates between patients with and without sepsis, and Phase III studies, which evaluate the test's performance in patients suspected to have the disorder. Performance characteristics, including sensitivity and specificity, positive and negative likelihood ratios, diagnostic odds ratio (the ratio of the odds of a positive test in a patient with sepsis compared with the odds of a positive test in a patient without sepsis), and summary receiver operator characteristic (SROC) curves were determined for these separate groups and then for all 18 studies combined.
In the 14 Phase II studies, procalcitonin had a positive likelihood ratio of 3.03, a negative likelihood ratio of 0.43, a diagnostic odds ratio of 7.79 (values greater than 100 indicate high accuracy while values less than 25 point to a lack of utility for the test), and an area under the curve of 0.79. Statistical analysis could not be performed on the 4 Phase III studies due to a high level of statistical heterogeneity resulting from variations in sample sizes; the larger studies in this group did trend toward a lack of diagnostic accuracy. When statistical analysis was performed on all 18 studies combined, test performance remained low; the mean values for sensitivity and specificity were 71% (95% confidence interval 67 to 76%), and the area under the SROC curve was only 0.78.
The study by Tang et al should give pause to clinicians who are considering adding procalcitonin measurements to their diagnostic protocols in the ICU. In a meta-analysis that appears methodologically sound, they demonstrate that the test characteristics for procalcitonin are not of sufficient quality to warrant reliance on this test in critically ill patients. In addition to the low mean sensitivity and specificity and somewhat low area under the ROC curve across the 18 studies included in the meta-analysis, the low positive and negative likelihood ratios are a concern.
The clinician standing at the bedside will not be looking for help from a diagnostic standpoint when his or her pre-test probability of the patient having sepsis is on the very low or very high side. Instead, such a clinician needs assistance when the pre-test probability is in the intermediate range. However, with a positive likelihood of 3 and a negative likelihood ratio of 0.43, the procalcitonin measurement is unlikely to change the pre-test probability enough to alter one's decisions as to whether or not to start antibiotics.
Another question to consider in deciding whether to incorporate this test into diagnostic protocols is how well the test performs relative to our current strategies. Among the studies in this meta-analysis, very few of them actually examined how the procalcitonin test performed when compared to a clinician's bedside assessment. Even if the test has a good sensitivity, specificity, and other test characteristics, it is not clear that we need to use it if it does not perform any better than clinicians already perform in its absence. For example, it might be worthwhile to know whether procalcitonin-based decision strategies decrease the inappropriate use of antibiotics, much the same way that bronchoalveolar lavage in the diagnosis of ventilator-associated pneumonia was shown to lead to more rational antibiotic use.
Until such data becomes available, or consistent data emerges from large studies on broad ranges of critically ill patients with better test characteristics than those found by Tang et al, clinicians should avoid using procalcitonin in the diagnosis of sepsis.