SHORT COMMUNICATION

Procalcitonin fails to predict bacteremia in SIRS patients: a cohort study M. Hoenigl,1,2 R. B. Raggam,3 J. Wagner,1 F. Prueller,3 A. J. Grisold,4 E. Leitner,4 K. Seeber,1 J. Prattes,1 T. Valentin,1 I. Zollner-Schwetz,1 G. Schilcher,5 R. Krause1 Linked Comment: www.youtube.com/IJCPeditorial

1

Section of Infectious Diseases and Tropical Medicine, Department of Internal Medicine, Medical University of Graz, Graz, Austria 2 Division of Pulmonology, Department of Internal Medicine, Medical University of Graz, Graz, Austria 3 Clinical Institute of Medical and Chemical Laboratory Diagnostics, Medical University of Graz, Graz, Austria 4 Institute of Hygiene, Microbiology and Environmental Medicine, Medical University of Graz, Graz, Austria 5 Division of Nephrology, Department of Internal Medicine, Medical University of Graz, Graz, Austria Correspondence to: Martin Hoenigl, Section of Infectious Diseases and Tropical Medicine, Medical University of Graz, Auenbruggerplatz 15, A8036, Graz Tel.: + 4331638581319 Fax: + 4331638514622 Email: martin.hoenigl@medunigraz. at and Robert Krause, Section of Infectious Diseases and Tropical Medicine, Medical University of Graz, Auenbruggerplatz 15, A-8036 Graz Tel.: + 4331638581796 Fax: + 4331638514622 Email: robert.krause@medunigraz. at

Disclosures All authors declare that they have no conflicts of interest. I hereby verify that all authors had access to the data and a role in writing the manuscript.

1278

SUMMARY

What’s known

Background: Procalcitonin (PCT) has previously been proposed as useful marker to rule out bloodstream-infection (BSI). The objective of this study was to evaluate the sensitivity of different PCT cut-offs for prediction of BSI in patients with community (CA)- and hospital-acquired (HA)-BSI. Methods: A total of 898 patients fulfilling systemic-inflammatory-response-syndrome (SIRS) criteria were enrolled in this prospective cohort study at the Medical University of Graz, Austria. Of those 666 patients had positive blood cultures (282 CA-BSI, 384 HA-BSI, enrolled between January 2011 and December 2012) and 232 negative blood cultures (enrolled between January 2011 and July 2011 at the emergency department). Blood samples for determination of laboratory infection markers (e.g. PCT) were collected simultaneously with blood cultures. Results: Procalcitonin was significantly (p < 0.001) higher in SIRS patients with bacteremia/fungemia than in those without. Receiver operating characteristic curve analysis revealed an area under the curve (AUC) value of 0.675 for PCT (95% CI 0.636–0.714) for differentiating patients with BSI from those without. AUC for IL-6 was 0.558 (95% CI 0.515–0.600). However, even at the lowest cutoff evaluated (i.e. 0.1 ng/ml) PCT failed to predict BSI in 7% (n = 46) of patients. In the group of patients with SIRS and negative blood culture 79% (n = 185) had PCT levels > 0.1. Conclusion: Procalcitonin was significantly higher in patients with BSI than in those without and superior to IL-6 and CRP. The clinical importance of this is questionable, because a suitable PCT threshold for excluding BSI was not established. An approach where blood cultures are guided by PCT only can therefore not be recommended.

Early diagnosis and appropriate treatment of bloodstream-infection (BSI) are the key factors in order to increase survival. Procalcitonin (PCT) has previously been proposed in numerous studies as useful marker to guide antimicrobial therapy in patients hospitalised for community-acquired (CA) infections. Some smaller studies have also suggested that PCT might be useful to rule out BSI in the emergency department and therefore may guide bloodcultures.

What’s new We found that even at the lowest cut-off evaluated (i.e. 0.1 ng/ml) PCT could not predict BSI in 7% (n = 46) of patients (30 patients with hospitalacquired and 16 with CA-BSI). In the group of patients with SIRS and negative blood culture 79% (n = 185) had PCT levels > 0.1. Driven from the results of our study cohort an approach where blood cultures are guided by PCT only can therefore not be recommended.

Introduction

Patients and methods

Despite advances in therapy and supportive care, bloodstream-infection (BSI) still represents a major cause of morbidity and mortality (1). Early diagnosis and appropriate treatment of BSI are the key factors in order to increase survival (2). Procalcitonin (PCT) has recently been proposed as useful marker to rule out BSI in the emergency department and to guide antimicrobial therapy in patients hospitalised for community-acquired (CA) infections (3–5). The objective of this study was to evaluate the sensitivity of different PCT cut-offs for prediction of bacteremia in systemic-inflammatory-response-syndrome (SIRS) patients with CA and those with hospitalacquired (HA)-BSI.

A total of 898 patients (all adults above 18 years of age) fulfilling SIRS criteria as described previously (6) were enrolled in this prospective cohort study at the Medical University of Graz, Austria. Of those 666 patients had positive blood cultures and were enrolled between January 2011 and December 2012. In 282 of the included 666 patients, the positive blood cultures were collected at the emergency department (not admitted during the last 5 days before presentation). These patients were therefore assigned to the CA-BSI group. Another 384 patients had been hospitalised for at least 48 h at the time the positive blood cultures were collected. These patients were therefore assigned to the HA-BSI group. Of 898 patients, 232 had negative blood cul-

ª 2014 John Wiley & Sons Ltd Int J Clin Pract, October 2014, 68, 10, 1278–1281. doi: 10.1111/ijcp.12474

Procalcitonin: useful to rule out blood stream infection?

tures and were enrolled between January 2011 and July 2011. The group consisted of patients that had presented with SIRS with subsequent collection of blood cultures at the emergency department. All these patients had not received antibacterial therapy at least 5 days prior to collection of the blood cultures and subsequently had negative blood culture results. Three pairs of blood cultures per patient were collected simultaneously, placed in the BACTECs blood culture systems (Becton Dickinson, Cockeysville, MD) and incubated for a maximum of 5 days as described previously (7). If coagulase negative staphylococci (CoNS) or gram-positive rods were detected two or more positive blood culture bottles from a blood culture set drawn by one venous puncture growing the same organism were required for a case to count as true bacteremia; cases that did not meet this requirements were excluded because of possible contamination. For the remaining bacterial/fungal pathogens one positive blood culture bottle was defined as sufficient evidence of bacteremia/fungemia. Blood samples for determination of laboratory infection markers were collected simultaneously with the initial blood cultures. For determination of PCT, the high sensitive Elecsys BRAHMS PCT test kit (ThermoFisher, Henningsdorf, Germany), with a detection limit of 0.02 ng/ml (analytical sensitivity) and a test linearity ranging from 0.02 to 100 ng/ml (analytical measuring range) on a Cobas 8000 system (Roche Diagnostics, Rotkreuz, Switzerland) was used. PCT values below 0.02 ng/ml were given as < 0.02 ng/ml and values exceeding > 100 ng/ml were diluted 1:4. Directly after patient enrolment determination of C-reactive protein (CRP), Interleukin (IL)6, and creatinine (from serum samples) was performed on the fully automated analyzer Cobas 8000 system (reagents all Roche Diagnostics). For calculation of sensitivities for BSI detection PCT cut-offs > 0.1, > 0.4 and > 0.5 ng/ml and IL-6 cut-off >10 pg/ml were used. Cut-offs were chosen consistent with previously published literature (3,4,7,8). Statistical analysis was performed using SPSS, version 20 (SPSS Inc., Chicago, IL). Patient groups were compared using Mann–Whitney U test. The pvalues of the Mann–Whitney U tests were not corrected for multiple comparisons and are therefore only descriptive. Receiver operating characteristics (ROC) curve analysis was performed for PCT and IL-6. Area under the curve (AUC) values were displayed including 95% confidence interval (CI). Correlation between PCT and duration of hospitalisation was calculated using Spearman-Rho correlation analysis. The study protocol was approved by the ª 2014 John Wiley & Sons Ltd Int J Clin Pract, October 2014, 68, 10, 1278–1281

local ethics committee, Medical University Graz, Austria (Ethical committee EC-number 21-469 ex 09/10; Clin.Trials.gov. number NCT01359891).

Results Demographical data as well as laboratory parameters in SIRS patients with BSI and those without are depicted in Table 1. While PCT and IL-6 were significantly higher in patients with BSI than in SIRS patients without, no differences were found for CRP and creatinine. ROC curve analysis revealed an AUC value of 0.675 for PCT (95% CI 0.636–0.714) for differentiating patients with BSI from those without. AUC for IL-6 was 0.558 (95% CI 0.515–0.600). Overall, 263/666 patients had bacteremia caused by Staphylococcus aureus (n = 82) or Escherichia coli (n = 181). Median PCT for these 263 patients was 1.64 ng/ml (IQR 0.42–10.87) and therefore markedly higher than in 83 patients with bacteremia caused by CoNS (median 0.37, IQR 0.16–0.95). ROC curve analysis for PCT to predict S. aureus or E. coli bacteremia (vs. negative blood culture results) revealed an AUC of 0.719 (95% CI 0.675–0.764; AUC for IL-6 0.601) while AUC for prediction of CoNS bacteremia was 0.507. While no significant correlation was found between PCT and duration of hospitalisation (p = 0.192, Spearman-Rho) in the overall study collective, there was a significant correlation in the subgroups of CA-BSI (p < 0.001, Spearman-Rho, higher PCT correlated with longer hospitalisation) and HA-BSI (p = 0.001). In the latter higher PCT (> 0.5 ng/ml) correlated with significantly shorter duration of hospitalisation (p = 0.021), since many of these patients (high PCT and HA-BSI) died, while no difference was found when using the 0.1 ng/ml PCT cut-off. Sensitivity of PCT for prediction of BSI (cut-off > 0.5 ng/ml) was 64% (CA-BSI 70%, HA-BSI 60%; BSI caused by S. aureus or E. coli 68.8%). The sensitivity increased slightly at a cut-off > 0.4 ng/ml (69%; CA-BSI 76%, HA-BSI 64%) and markedly at > 0.1 ng/ml (93%; CA-BSI 94%, HA-BSI 92%; BSI caused by S. aureus or E. coli 94.7%). At the time of blood culture sampling, PCT was ≤ 0.1 ng/ml in a total of 46 patients with BSI (31 males, 15 females, median age 63 years, median duration of hospitalisation 11 days). Sixteen of these 46 patients had CA-BSI. Detected pathogens included CoNS (n = 8), E. coli (n = 3), S. aureus (n = 3), and others (n = 2). Two of these 16 patients had to be admitted to the ICU, one patient died because of septic shock within 24 h, all others survived at day 90 after admission. The remaining 30/46 patients had

1279

1280

Procalcitonin: useful to rule out blood stream infection?

Table 1 Demographical data, duration of hospitalisation and laboratory parameters (medians and IQRs displayed) of

SIRS patients with BSI and those without

Age (years) Sex (f/m) Procalcitonin (ng/ml) Interleukin-6 (pg/ml) C-reactive protein (mg/l) Serum creatinine (mg/dl) Days hospitalised after collection of positive blood culture (days)

BSI (n = 666)*

Negative blood cultures (n = 232)*

66 (IQR 54–76) 284/382 1.06 (IQR 0.31–7.6) 223 (IQR 82–840) 108 (IQR 50–201) 1.25 (IQR 0.89–2.30) 13 (IQR 8–21)

68 (IQR 51–77) 95/137 0.29 (IQR 0.12–1.30) 174 (76–401) 108 (IQR 40–210) 1.16 (IQR 0.91–1.74) 8 (IQR 5–11)

p-value (if significant)

< 0.001 0.01

< 0.001

BSI, blood stream infection; f, female; IQR, inter quartile range; m, male. *All data available from all included patients.

HA-BSI caused by S. aureus (n = 6), Enterococcus spp. (n = 5), Pseudomonas spp. (n = 4), CoNS (n = 4), E. coli, Streptococcus spp., Klebsiella pneumoniae, Candida spp. (n = 2 each), and others (n = 3). In three out of these 30 patients blood cultures were drawn at ICU while two had to be admitted to ICU because of septic shock within 3 days after blood cultures had been drawn. All but two patients survived at day 90. IL-6 was above 10 pg/ml in 44/46 (96%) of patients with BSI and low PCT (≤ 0.1 ng/ml). Sensitivity for IL-6 (cut-off 10 pg/ml) for prediction of BSI was 99.6%. Three patients with BSI had IL-6 levels ≤ 10 pg/ml. With regard to patients with SIRS and negative blood culture result (n = 233, had not received antibacterial therapy at least 5 days prior to collection of the blood cultures), 185 (79%) had PCT levels > 0.1 ng/ml and 227 (97%) IL-6 levels > 10 ng/ml.

Discussion We evaluated PCT for prediction of BSI in 666 patients with BSI and found that PCT was significantly higher in patients with bacteremia/fungemia than in those without and proved to be superior to IL-6 and CRP. However, even at the lowest cut-off evaluated (i.e. > 0.1 ng/ml) PCT could not predict BSI in 7% (n = 46) of patients. A biomarker that convincingly rules out BSI0 s might be useful to guide blood culturing and select patients that may benefit from blood cultures. PCT has been proposed by a number of studies with smaller patient collectives to have the potential of ruling out BSI in the emergency department and in patients hospitalised for CA-infections (3–5,8). While in one study, a serum PCT level of < 0.4 ng/ml has been reported as an accurate cut-off to rule out BSI in 22 patients hospitalised with CA-BSI (4), recent

studies have proposed a distinctive lower cut-off of 0.1 ng/ml at emergency departments (3,8). Also in the latter studies, however, the number of bacteremic patients was small. With regard to the poor specificity of the currently proposed cut-offs our results are in accordance with previously published studies (7,8). This study has some limitations which mostly relate to the blood culture negative cohort (n = 232) that had been enrolled during the first months of the study only. Reliable calculation of positive and negative predictive values was therefore not possible. Nevertheless, our results clearly point out that PCT is a promising marker for prediction of BSI. However, our data also indicate that PCT – even at a very low cut-off – may not only fail to predict BSI in one of 15 patients, but may also be an unspecific predictor of BSI resulting positive in about 80% of patients with clinical suspicion of BSI but subsequent negative blood culture results. At our institution, a PCT guided approach for the decision whether to obtain blood cultures or not (with a PCT cut-off of 0.1 ng/ml) would have resulted in a 20% decrease in blood cultures obtained unnecessarily in patients without BSI, but conversely would have missed 7% of patients with BSIs. Considering (i) the potential of prolonged hospitalisations, morbidity and mortality because of delayed diagnosis in BSI patients with low PCT and (ii) the fact that PCT may reduce negative blood cultures by 20% only, an approach for blood cultures guided by PCT only may not seem feasible.

Acknowledgements The study was conducted as part of our routine work. No extra funding obtained. ª 2014 John Wiley & Sons Ltd Int J Clin Pract, October 2014, 68, 10, 1278–1281

Procalcitonin: useful to rule out blood stream infection?

Author contributions Martin Hoenigl, Robert Krause, and Reinhard B Raggam concepted and designed the study, analysed the data, drafted the manuscript and supervised Jasmin Wagner, Katharina Seeber and J€ urgen Prattes. Jasmin Wagner, J€ urgen Prattes, Gernot Schilcher and Martin Hoenigl collected the extensive laboratory and clinical data. Reinhard Raggam and Florian Prueller were responsible for measurement of biomarkers in the

References 1 Sogaard M, Norgaard M, Dethlefsen C, Schonheyder HC. Temporal changes in the incidence and 30-day mortality associated with bacteremia in hospitalized patients from 1992 through 2006: a population-based cohort study. Clin Infect Dis 2011; 52(1): 61–9. 2 Jeanrenaud P, Hammell C, Dempsey GA et al. Markers of bacterial infection in the critically ill: a comparison of procalcitonin, C reactive protein and the neutrophil band count. J Infect 2012; 64(5): 540–2. 3 Albrich WC, Mueller B. Predicting bacteremia by procalcitonin levels in patients evaluated for sepsis

ª 2014 John Wiley & Sons Ltd Int J Clin Pract, October 2014, 68, 10, 1278–1281

1281

laboratory. Andrea J Grisold and Eva Leitner contributed cases of bacteremia diagnosed at the Institute of Hygiene. Jasmin Wagner, Ines Zollner-Schwetz and Thomas Valentin made major contributions in interpretation and analysis of the data and critically revised of the intellectual content together with Florian Prueller, Andrea J Grisold, Eva Leitner, Katharina Seeber, Gernot Schilcher and J€ urgen Prattes. All contributing authors approved the final version to be published.

in the emergency department. Expert Rev Anti Infect Ther 2011; 9(6): 653–6. 4 Chirouze C, Schuhmacher H, Rabaud C et al. Low serum procalcitonin level accurately predicts the absence of bacteremia in adult patients with acute fever. Clin Infect Dis 2002; 35(2): 156–61. 5 Lee SH, Chan RC, Wu JY, Chen HW, Chang SS, Lee CC. Diagnostic value of procalcitonin for bacterial infection in elderly patients – a systemic review and meta-analysis. Int J Clin Pract 2013; 67(12): 1350–7. 6 Bone RC, Balk RA, Cerra FB et al. Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. the ACCP/SCCM consensus conference committee. American College of

Chest Physicians/Society of Critical Care Medicine, 1992. Chest 2009; 136(5 Suppl.): e28. 7 Hoenigl M, Raggam RB, Wagner J et al. Diagnostic accuracy of soluble urokinase plasminogen activator receptor (suPAR) for prediction of bacteremia in patients with systemic inflammatory response syndrome. Clin Biochem 2013; 46(3): 225–9. 8 Riedel S, Melendez JH, An AT, Rosenbaum JE, Zenilman JM. Procalcitonin as a marker for the detection of bacteremia and sepsis in the emergency department. Am J Clin Pathol 2011; 135(2): 182–9.

Paper received December 2013, accepted May 2014