0
Editorial |

Additional Real-World Evidence Supporting Procalcitonin as an Effective Tool to Improve Antibiotic Management and Cost of the Critically Ill Patient FREE TO VIEW

Philipp Schuetz, MD, MPH; Peter M. Wahl, ScD
Author and Funding Information

FINANCIAL/NONFINANCIAL DISCLOSURES: The authors has reported to CHEST the following: P. S. is supported by the Swiss National Science Foundation (SNSF Professorship, PP00P3_150531) and the Forschungsrat of the Kantonsspital Aarau (1410.000.058 and 1410.000.044). P. S. received research support from BRAHMS/Thermo Fisher, bioMérieux, Roche, Abbott, Nestlé, and Novo Nordisk. P. M. W. is a full-time employee of Covance, Inc. The views expressed by P. M. W. herein are those of the author and do not necessarily reflect the views of Covance, Inc.

aUniversity Department of Medicine, Kantonsspital Aarau and Medical Faculty of the University of Basel, Basel, Switzerland

bCovance, Inc, Princeton, NJ

CORRESPONDENCE TO: Philipp Schuetz, MD, MPH, University Department of Medicine, Kantonsspital Aarau, Tellstrasse, CH-5001 Aarau, Switzerland


Copyright 2016, American College of Chest Physicians. All Rights Reserved.


Chest. 2017;151(1):6-8. doi:10.1016/j.chest.2016.07.014
Text Size: A A A
Published online

Traditionally, studies investigating the usefulness of a biomarker focus on diagnostic measures such as sensitivity and specificity. This approach, however, mandates the existence of a well-accepted reference standard. For biomarkers that are used to help treat patients with systemic infections and sepsis there is no such reference standard, with blood cultures having low sensitivity of only 10% to 30%. Thus, randomized controlled trials are needed to assess the benefits and limitations of infection biomarkers by comparing outcomes of marker-assessed patients with patients receiving routine care. In the case of procalcitonin (PCT) and its effect in the treatment of patients with sepsis, numerous studies have investigated how well this marker differentiates patients with true sepsis from patients presenting with a sepsis-like syndrome but no infectious etiology. Depending on the cutoff used, reported sensitivities and specificities range between 70% and 95%, with the lack of a reference standard making the interpretation of these results challenging. Importantly, several randomized trials have investigated the effects of PCT protocols and report important reductions in antibiotic use in the range of 30% to 70%, depending on the clinical setting and main infection diagnosis,, with a recent trial finding a significant survival benefit associated with the use of PCT in the critical care setting.

FOR RELATED ARTICLE SEE PAGE 23

Still, physicians in randomized trials may behave differently than in typical care settings, where clinical protocol compliance rates may be lower because they know that they are not being watched (commonly referred to as the “Hawthorne effect”). This may influence both the intervention and the control groups and thus somewhat mask the true effect a biomarker may have on the treatment of patients. It is therefore important to also investigate so-called real-world data—usually by studying public databases or patient registries—to further broaden and expand findings from randomized trials to usual care.

Given the promising results from randomized trials, it is important to know how PCT impacts the clinical management of patients in real-world settings. Although these observational data sets are not randomized and thus bear the risk to internal validity of confounding by indication, there are several statistical approaches—such as propensity score matching—that help to lower that risk.

Balk et al, writing in this issue of CHEST, provide an important piece of information by investigating more than 33,000 patients treated with PCT in the critical care unit and almost 100,000 critically ill patients without PCT treatment, all derived from the Premier Healthcare Database. To account for differences in the two populations, the authors used a 1:3 propensity score match and limited their analysis to patients with a PCT determination within the first day of ICU admission. This approach addresses potential confounding by important factors such as patient demographics, hospital characteristics (urbanicity, teaching status, bed size, region), and patient clinical characteristics (admission type, admission source, number of types of antibiotics received on the first ICU day, dialysis on or before the first ICU day, ventilator use on or before the first ICU day, use of vasopressors or inotropes on the first ICU day, prior hospitalization within 30 days, and presence of the 10 most frequent admission diagnoses).

The propensity score is the estimated probability of the patient receiving the treatment of interest (PCT testing) relative to the comparator treatment (no PCT testing), conditional on covariates measured at baseline. The goal of propensity score-based methods in the setting of observational studies is to eliminate systematic bias in effect estimates caused by measured factors that are associated with both treatment choice and outcome (ie, baseline confounders). In theory, propensity score matching accomplishes this by achieving balance between comparison groups in the measured baseline confounders. This method relies on a few assumptions to facilitate causal inference, including that the propensity score and outcome models are both properly specified, that all baseline confounders have been measured, and that all confounder levels (eg, male and female subjects) are represented by members of each treatment group (ie, positivity).

Propensity score matching ensures that the latter is the case by matching exposed with unexposed patients within a prespecified distance between their estimated propensity scores. The authors observed significant overlap in the distribution of propensity scores between the two treatment groups, and therefore were arguably justified in targeting a causal estimate of PCT administration on the outcomes of interest by matching patients receiving PCT with their nearest non-PCT propensity score neighbor. Assuming the authors properly measured the strongest confounders, matching on the propensity score enhances the internal validity of the study, although readers cannot decide for themselves the degree to which this was accomplished since results from crude (non-propensity score-matched) analyses were not shown. Residual confounding is likely present due to some of the limitations cited by the authors, as well as the fact that the study data set does not capture preadmission comorbid conditions recorded by non-Premier practices or institutions.

Balk et al may also have conditioned on post-treatment (ie, post-PCT) characteristics by including factors in their propensity score model that may have occurred after PCT was undertaken (eg, antibiotics received on the first ICU day). Including measures of post-treatment patient characteristics in the propensity score model may result in effectively estimating direct effects of treatment and thus shift the marginal treatment effect estimate toward the null, as evidenced by the results displayed in Table 3 in Balk et al (compared with those in Table 2). The more onerous effect of conditioning on post-treatment factors is creating selection—or collider—bias, a systematic bias of the effect estimate induced analytically by conditioning on a factor that shares a common cause (a baseline factor or treatment) with the outcome(s) of interest. Nevertheless, estimating causal effects from observational studies conducted with administrative hospital data is challenging, and the authors should be commended for employing methods at their disposal to mitigate potential confounding at baseline.

Despite these concerns the results were quite impressive, with PCT-treated patients showing significant reductions in total antibiotic exposure (16.2 vs 16.9 days), and in total and ICU lengths of stay (11.6 vs 12.7 and 5.1 vs 5.3 days, respectively), resulting in an approximately 10% reduction in hospital costs ($30,454 vs $33,213). Although mortality was not improved, patients treated according to the PCT protocol were more likely to be discharged home. The results of this study also contribute to current knowledge as data from US patients have been scarce, with most interventional research being conducted in Europe and Asia.

Current Surviving Sepsis Campaign guidelines provide, at present, only a grade 2C recommendation for the use of PCT testing and “suggest the use of low procalcitonin…to assist the clinician in the discontinuation of empiric antibiotics in patients who…have no subsequent evidence of infection.(p172), Results of the recent randomized SAPS (Stop Antibiotics on Procalcitonin Guidance Study) trial in conjunction with real-life data reported by Balk et al in this issue of CHEST are convincing and should lead physicians to more widespread use of PCT protocols for the treatment of patients in the critical care setting.

References

Schuetz P. .Aujesky D. .Müller C. .Müller B. . Biomarker-guided personalised emergency medicine for all—hope for another hype? Swiss Med Wkly. 2015;145:w14079- [PubMed]journal. [PubMed]
 
Wacker C. .Prkno A. .Brunkhorst F.M. .Schlattmann P. . Procalcitonin as a diagnostic marker for sepsis: a systematic review and meta-analysis. Lancet Infect Dis. 2013;13:426-435 [PubMed]journal. [CrossRef] [PubMed]
 
Schuetz P. .Briel M. .Christ-Crain M. .et al Procalcitonin to guide initiation and duration of antibiotic treatment in acute respiratory infections: an individual patient data meta-analysis. Clin Infect Dis. 2012;55:651-662 [PubMed]journal. [CrossRef] [PubMed]
 
Schuetz P. .Amin D.N. .Greenwald J.L. . Role of procalcitonin in managing adult patients with respiratory tract infections. Chest. 2012;141:1063-1073 [PubMed]journal. [CrossRef] [PubMed]
 
de Jong E. .van Oers J.A. .Beishuizen A. .et al Efficacy and safety of procalcitonin guidance in reducing the duration of antibiotic treatment in critically ill patients: a randomised, controlled, open-label trial. Lancet Infect Dis. 2016;16:819-827 [PubMed]journal. [CrossRef] [PubMed]
 
Albrich W.C. .Dusemund F. .Bucher B. .et al Effectiveness and safety of procalcitonin-guided antibiotic therapy in lower respiratory tract infections in “real life”: an international, multicenter poststudy survey (ProREAL). Arch Intern Med. 2012;172:715-722 [PubMed]journal. [CrossRef] [PubMed]
 
Balk R.A. .Kadri S.S. .Cao Z. .Robinson S.B. .Lipkin C. .Bozzette S.A. . Effect of procalcitonin testing on health-care utilization and costs in critically ill patients in the United States. Chest. 2017;151:23-33 [PubMed]journal
 
Hernan M.A. .Robins J.M. . Estimating causal effects from epidemiological data. J Epidemiol Community Health. 2006;60:578-586 [PubMed]journal. [CrossRef] [PubMed]
 
Pearl J. Direct and indirect effects. In:Proceedings of the Seventeenth Conference on Uncertainty in Artificial Intelligence[2001 Aug 2-5, Seattle, WA]. San Francisco, CA: Morgan Kaufmann; 2001:411-420.
 
Hernan M.A. .Hernandez-Diaz S. .Robins J.M. . A structural approach to selection bias. Epidemiology. 2004;15:615-625 [PubMed]journal. [CrossRef] [PubMed]
 
Dellinger R.P. .Levy M.M. .Rhodes A. .et al Surviving Sepsis Campaign: international guidelines for management of severe sepsis and septic shock, 2012. Intensive Care Med. 2013;39:165-228 [PubMed]journal. [CrossRef] [PubMed]
 
Schuetz P. .Müeller B. . Procalcitonin in critically ill patients: time to change guidelines and antibiotic use in practice. Lancet Infect Dis. 2016;16:758-760 [PubMed]journal. [CrossRef] [PubMed]
 

Figures

Tables

References

Schuetz P. .Aujesky D. .Müller C. .Müller B. . Biomarker-guided personalised emergency medicine for all—hope for another hype? Swiss Med Wkly. 2015;145:w14079- [PubMed]journal. [PubMed]
 
Wacker C. .Prkno A. .Brunkhorst F.M. .Schlattmann P. . Procalcitonin as a diagnostic marker for sepsis: a systematic review and meta-analysis. Lancet Infect Dis. 2013;13:426-435 [PubMed]journal. [CrossRef] [PubMed]
 
Schuetz P. .Briel M. .Christ-Crain M. .et al Procalcitonin to guide initiation and duration of antibiotic treatment in acute respiratory infections: an individual patient data meta-analysis. Clin Infect Dis. 2012;55:651-662 [PubMed]journal. [CrossRef] [PubMed]
 
Schuetz P. .Amin D.N. .Greenwald J.L. . Role of procalcitonin in managing adult patients with respiratory tract infections. Chest. 2012;141:1063-1073 [PubMed]journal. [CrossRef] [PubMed]
 
de Jong E. .van Oers J.A. .Beishuizen A. .et al Efficacy and safety of procalcitonin guidance in reducing the duration of antibiotic treatment in critically ill patients: a randomised, controlled, open-label trial. Lancet Infect Dis. 2016;16:819-827 [PubMed]journal. [CrossRef] [PubMed]
 
Albrich W.C. .Dusemund F. .Bucher B. .et al Effectiveness and safety of procalcitonin-guided antibiotic therapy in lower respiratory tract infections in “real life”: an international, multicenter poststudy survey (ProREAL). Arch Intern Med. 2012;172:715-722 [PubMed]journal. [CrossRef] [PubMed]
 
Balk R.A. .Kadri S.S. .Cao Z. .Robinson S.B. .Lipkin C. .Bozzette S.A. . Effect of procalcitonin testing on health-care utilization and costs in critically ill patients in the United States. Chest. 2017;151:23-33 [PubMed]journal
 
Hernan M.A. .Robins J.M. . Estimating causal effects from epidemiological data. J Epidemiol Community Health. 2006;60:578-586 [PubMed]journal. [CrossRef] [PubMed]
 
Pearl J. Direct and indirect effects. In:Proceedings of the Seventeenth Conference on Uncertainty in Artificial Intelligence[2001 Aug 2-5, Seattle, WA]. San Francisco, CA: Morgan Kaufmann; 2001:411-420.
 
Hernan M.A. .Hernandez-Diaz S. .Robins J.M. . A structural approach to selection bias. Epidemiology. 2004;15:615-625 [PubMed]journal. [CrossRef] [PubMed]
 
Dellinger R.P. .Levy M.M. .Rhodes A. .et al Surviving Sepsis Campaign: international guidelines for management of severe sepsis and septic shock, 2012. Intensive Care Med. 2013;39:165-228 [PubMed]journal. [CrossRef] [PubMed]
 
Schuetz P. .Müeller B. . Procalcitonin in critically ill patients: time to change guidelines and antibiotic use in practice. Lancet Infect Dis. 2016;16:758-760 [PubMed]journal. [CrossRef] [PubMed]
 
NOTE:
Citing articles are presented as examples only. In non-demo SCM6 implementation, integration with CrossRef’s "Cited By" API will populate this tab (http://www.crossref.org/citedby.html).

Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging & repositioning the boxes below.

  • CHEST Journal
    Print ISSN: 0012-3692
    Online ISSN: 1931-3543