Procalcitonin levels predict to identify bacterial strains in blood cultures of septic patients Takao Arai MD, PhD, Shoichi Ohta MD, PhD, Junya Tsurukiri MD, PhD, Kenichiro Kumasaka MD, Katsuhiro Nagata MD, Taihei Okita MD, Taishi Oomura MD, Akira Hoshiai MD, Masaharu Koyama PhD, Tetsuo Yukioka MD, PhD PII: DOI: Reference:
S0735-6757(16)30495-8 doi: 10.1016/j.ajem.2016.08.009 YAJEM 56044
To appear in:
American Journal of Emergency Medicine
Received date: Revised date: Accepted date:
14 April 2016 4 August 2016 5 August 2016
Please cite this article as: Arai Takao, Ohta Shoichi, Tsurukiri Junya, Kumasaka Kenichiro, Nagata Katsuhiro, Okita Taihei, Oomura Taishi, Hoshiai Akira, Koyama Masaharu, Yukioka Tetsuo, Procalcitonin levels predict to identify bacterial strains in blood cultures of septic patients, American Journal of Emergency Medicine (2016), doi: 10.1016/j.ajem.2016.08.009
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
TITLE:
T
Procalcitonin levels predict to identify bacterial strains in blood
Short running head:
MA NU
Bacterial infection predicted by procalcitonin
SC
RI P
cultures of septic patients
Authors:
Takao Arai*, MD, PhD, Shoichi Ohta, MD, PhD, Junya Tsurukiri, MD, PhD, Kenichiro
ED
Kumasaka, MD, Katsuhiro Nagata, MD, Taihei Okita, MD, Taishi Oomura, MD, Akira
PT
Hoshiai, MD, Masaharu Koyama, PhD, Tetsuo Yukioka, MD, PhD
Department of Emergency and Critical Care Medicine, Trauma and Emergency Center
CE
Hachioji Medical Center of Tokyo Medical University
*
AC
1163 Tatemachi, Hachioji-shi, Tokyo 193-0998, Japan
Corresponding author:
Department of Emergency and Critical Care Medicine, Trauma and Emergency Center Hachioji Medical Center of Tokyo Medical University 1163 Tatemachi, Hachioji-shi, Tokyo 193-0998, Japan Tel.: +81-42-665-5611 E-mail:
[email protected]
ACCEPTED MANUSCRIPT
INTRODUCTION
T
Rapid and accurate diagnosis of sepsis would be clinically useful as this can
RI P
provide guidance in early appropriate antibiotic treatment and improve patient survival. Although blood cultures (BCs) are considered as the “gold standard” for the isolation
SC
and identification of causative pathogens from patients with sepsis [1], this method remains insufficiently time critical and cannot provide assistance with the early
MA NU
therapeutic decisions of emergency department (ED) physicians. Furthermore, the BC results obtained in EDs have high false-positive rates caused by skin contaminant organisms, resulting in unnecessary diagnostic measures, hospitalization, and
ED
unwarranted antimicrobial therapy [2-4]. While the BC results have also false-negative problems in automated blood culture system [5] and falsely negative BCs occurs in 5%
PT
cases of infective endocarditis [6]. Therefore, as many as blood cultures may be needed to achieve a high detection rate [7, 8] within the allowance of status of patient, in a case
CE
of suspected bacteremia. Thus, given the slowness, low sensitivity, and specificity of
AC
BCs, more rapid and sensitive techniques are eagerly awaited. Recently, clinical laboratories have begun to move towards novel diagnostic
approaches. Indeed, molecular techniques such as polymerase chain reaction (PCR) and mass spectroscopy provide substantially more rapid and specific information on organism identification and the presence of resistance mechanism [9]. However, as for pathogen identification, a multipathogen probe-based real-time PCR is compatible to BCs in patients with suspected sepsis [10]. Furthermore, these methods are very expensive, time-consuming, and require special equipment. Although laboratory tests with high sensitivity and specificity for sepsis have not yet been established [11], the combination of readily available laboratory tests such
ACCEPTED MANUSCRIPT
as diagnostic tests using procalcitonin (PCT), C-reactive protein (CRP), and more effectively identify non-infectious from
infectious
T
interleukin-6 could
RI P
inflammations [12]. PCT was described as a marker of sepsis in 1993 [13]. Moreover, the induction of PCT is associated with the activation and adherence of monocytic cells,
SC
which occurs during sepsis as well as in other conditions such as after tissue trauma. Indeed, PCT has the highest sensitivity and specificity for predicting systemic
MA NU
inflammation compared with conventional serum markers [14-16]. Previous studies have reported that Gram-negative (G-) rods (GNR)-positive patients had a higher PCT value than Gram-positive (G+) cocci (GPC)-positive patients [17-20]. However,
ED
GPC-positive or GNR-positive patients have been difficult to predict by laboratory tests alone in EDs.
PT
We previously applied principal component analysis (PCA), one of the most widely used multivariate statistical techniques [21], to detect the degree of relational
CE
contribution of 4 laboratory tests to BC results [22], yielding 2 components. The score
AC
of component 1 was high when the PCT value was high and the platelet (PLT) value was low. In contrast, the score of component 2 was predominantly affected by the value of the white blood cell (WBC), the most common marker of systemic inflammation [22]. Thus, each component obtained from PCA appears to reflect an association with a clinical manifestation of ED patients. However, some patients suspected of having bacteremia predicted from a high PCT value and distributed in the first and fourth quadrants, were BC-negative. Some of those patients might have been given antibiotic treatment by a physician in another hospital before arriving at our hospital. This can make the results unclear. Therefore, we carefully designed this study to predict BC results of septic patients in ED, focusing on patients without prior antibiotic treatment.
ACCEPTED MANUSCRIPT
Furthermore, we attempted to identify differences between GPC and GNR using
AC
CE
PT
ED
MA NU
SC
RI P
T
multivariate statistical analyses.
ACCEPTED MANUSCRIPT
MATERIAL AND METHODS
T
Study Setting and Population
RI P
This retrospective study was conducted at the Department of Emergency and Critical Care Medicine at Tokyo Medical University Hachioji Medical Center, using
SC
data obtained from critically ill patients treated between January 2012 and June 2014. This study was approved by the Ethics Committee of Tokyo Medical University. All
MA NU
patients enrolled in the study provided written informed consent. All patients (18 years or older) with (1) a fever of more than 38ºC, (2) chills and shivering, or (3) suspected bacterial infection for sepsis, had blood samples taken
ED
for laboratory tests and BC. Blood samples from serum-separator tubes were centrifuged, and an aliquot of the serum was removed and stored at -70ºC for PCT
PT
analysis. Clinical data was abstracted from the medical records including age, sex, mortality, and principal diagnosis (International Statistical Classification of Diseases
CE
(ICD), 10th Revision). Patients who were administered antimicrobial therapy before
AC
arriving at our hospital were excluded from this study.
Laboratory Examinations Measurements of 4 sets of laboratory tests (PCT, CRP, PLT, and WBC) were conducted as previously reported [22]. BC results were obtained using BACTEC9120 (Japan Becton, Dickinson and Company, Tokyo, Japan). Bacteria from positive BCs were further identified using standard laboratory methods. Contamination was considered when any of the following bacteria were isolated from 1 of the 2 sets: Propionibacterium spp., Bacillus spp., Corynebacterium spp., coagulase-negative staphylococci, or Clostridium perfringens [23, 24].
ACCEPTED MANUSCRIPT
T
Statistical Analysis
RI P
Statistical analysis was conducted using SPSS version 22 (SPSS, Chicago, IL). Different categories were compared using Tukey’s multiple comparison test. Values
SC
were expressed as mean ± standard deviation (SD). Comparisons between proportions were made using the test. P < 0.05 was considered to indicate a statistically
MA NU
significant difference. PCA was applied after standardization as previously reported [22] to investigate the contribution of each test. In choosing the components, we fixed the
AC
CE
PT
ED
number of components to 2.
ACCEPTED MANUSCRIPT
RESULTS
T
A total of 315 patients were enrolled in a period of 30 months. The sex, mean
RI P
age, mortality, and principal diagnosis (ICD 10) are shown in Table 1. Approximately 62% of patients were men, their mean age was 65.4±19.5 years, and the diseases of the
SC
respiratory system was the greatest number (86 cases) among the diagnosis enrolled in this study (Table 1). Thirty-one patients who were administered antibiotics before
MA NU
arriving at our hospital and 113 patients who did not have all of the 4 sets of laboratory tests on the same date were excluded as explained in Material and Methods. A total of 171 patients with 4 sets of laboratory tests were analyzed by PCA.
ED
PCA for the loadings of the variables such as PCT, WBC, PLT, and CRP yielded 2 components (Table 2). The proportions of variance for principal component 1
PT
and principal component 2 were 41.4% and 29.2%, respectively, and 70.6% of information of the 4 parameters was compressed in 2 dimensions. Principal component
CE
1 which is produced by linear combinations with the highest PCT value and the lowest
AC
PLT value possibly indicates bacterial (fungal) infection, as the score of principal component 1 tends to be high when the PCT value is high and the PLT value is low. Principal component 2 which is produced by linear combinations with the highest WBC value and the lowest PCT value possibly indicates a systemic inflammation response. These descriptions were fairly consistent with our previous results [22]. The PCA values of components 1 and 2 were plotted in Fig. 1 identifying BC results as positive or negative (Fig. 1). The mean PCA values for component 1 between BC-negative patients and BC-positive patients were significantly different (positive vs negative: -0.31 ± 0.65, n = 124 vs 0.71 ±1.21, n = 47, p < 0.01). We arbitrary divided component 1 into 5 strata to analyze the BC-positive and
ACCEPTED MANUSCRIPT
BC-negative proportions (Table 3). All patients whose component 1 were less than – 1.0
T
had negative BC (n = 21). As component 1 increased up to 2.0, the BC-positive
RI P
proportions increased and all patients whose component 1 were more than 2.0 had positive BC (n = 8).
SC
Figure 2 shows only the patients whose BCs were positive, identifying whether their strains were GPC or GNR. For patients whose component 1 were more than 2.00,
MA NU
75% had GNR in the blood (n = 8) and among them for patients whose component 2 were more than -1.0, all of them were GNR-positive (n = 5). There were no GPC-positive patients found for those values. The mean value of component 1 of the
ED
GPC-positive patients (0.28 ± 0.87, n = 28) was significantly lower than that of the GNR-positive patients (1.58 ± 1.41, n = 19, p < 0.01). As shown in Table 4, we also
PT
divided component 1 into 5 strata to analyze the strains as either GPC or GNR. The GNR proportions were increased and the GPC proportions were decreased in the high
AC
CE
stratum of component 1.
ACCEPTED MANUSCRIPT
DISCUSSION
T
In this study, the values obtained from PCA of the 4 laboratory tests have
RI P
shown significant differences between BC-negative patients and GPC- or GNR-positive patients. Furthermore, PCA of the 4 laboratory tests showed a certain range in the first
SC
and fourth quadrants to which only the patients with GNR are distributed. These novel findings suggest that the predicted values applied by PCA potentially enables the
MA NU
identification of bacterial strains in BCs and the possibility of deciding the class of antibiotic treatments for patients distributed to this range in the first and fourth quadrants in the ED prior to obtaining the final BC results.
ED
Despite the exclusion of patients with prior antibiotic treatment, the scores we obtained for the loadings of the variables were similar to those we previously obtained
PT
[22] by applying 4 clinical biomarkers (4-dimensional data) to PCA to predict BC results. This indicates the reproducibility of this approach.
CE
In this study, 73% of the patients whose component 1 was more than 1.0 had
AC
positive BCs, and 69% of the patients whose component 1 was less than 1.0 had negative BCs. This implies that the patients distributed to more than 1.0 in component 1 require immediate BC if this has not yet been performed, and the patients distributed to less than 1.0 in component 1 do not require immediate attention. Thus, PCA using the 4 clinical biomarkers evaluated in this study may be a useful approach for predicting BC results. We also examined whether the differences obtained from the predicted values for BC results are useful for making a distinction between bacterial strains based on previous reports [17-20]. Although the proportions of BC-positive and BC-negative patients distributed in each quadrant were comparable with those of a previous study
ACCEPTED MANUSCRIPT
[22], subtle differences in the values between the previous study and the present study
T
were observed, which might be attributed to the previous administration of antibiotics.
RI P
In the case of that one is greater than 2.00 for component 1, this could be considered as BC-positive, whereas in the case of that one is greater than 2.00 for component 1 and
SC
-1.00 for component 2, this could be considered as GNR-positive. These imply that we may be able to have a reasonable speculation regarding the utilization of these
MA NU
laboratory test results in ED before obtaining the BC results.
Taken together, our results suggest a high proportion of GNR infection in cases wherein the PCT was high and the PLT value was low for component 1. Our new
ED
approach to the utilization of PCA for predicting bacterial infections in BC is consistent with previous reports [17-20], suggesting that GNR bacteria induce septic shock with
PT
lipopolysaccharide (LPS). LPS is the principal component of the outer leaflet of the outer membrane of GNR bacteria, and is recognized as the most potent microbial
CE
mediator implicated in the pathogenesis of sepsis sequelae and septic shock, resulting in
AC
an increase in the PCT value. Notably, Guo et al have recently reported that PCT was the only predictor of GNR BC as demonstrated by univariate and multivariate logistic regression analyses [25]. Based on the scoring method calculated from PCA, clinicians can strength the clinical decision whether the blood culture should be performed. Therefore, this might reduce contamination (false-positive rate). This scoring method calculated from PCA can also clarify the difference between GNR and GPC, that clinicians has evaluated based on Gram’s stain or the infected organ, from another aspect. The clinical significance of this scoring method calculated from PCA will be further increased when the sepsis of unknown source occurs. The study focused on these cases will be required
ACCEPTED MANUSCRIPT
in the future.
T
The present study has certain limitations. This is a single-center retrospective
RI P
study. Furthermore, although the blood cultures are regarded as significant, the blood culture has the false-positive and false-negative problems, affecting the reliability of the
SC
data.
Additional studies of different patient populations are anticipated to further
MA NU
clarify the usefulness of PCA for predicting bacterial infections and identifying bacterial strains in BCs of septic patients based on the values of laboratory test biomarkers
AC
CE
PT
ED
including PCT.
ACCEPTED MANUSCRIPT
CONCLUSION
T
This study showed that PCA applied to the combination of 4 commonly
RI P
performed laboratory tests, instead of PCT alone, potentially provides a useful tool for predicting BC results for sepsis diagnosis and further identification of bacterial strains
AC
CE
PT
ED
MA NU
SC
as either GNR or GPC.
ACCEPTED MANUSCRIPT
ACKNOWLEDGEMENT
T
The authors thank Dr. Edward F. Barroga, Associate Professor and Senior
AC
CE
PT
ED
MA NU
SC
Medical University for editing the English manuscript.
RI P
Medical Editor of the Department of International Medical Communications of Tokyo
ACCEPTED MANUSCRIPT
REFERENCES
RI P
T
[1] Peters RP,van Agtmael MA, Danner SA, Savelkoul PH, Vandenbroucke-Grauls CM. New developments in the diagnosis of bloodstream infections. Lancet Infect Dis. 2004;4(12):751-60
MA NU
SC
[2] 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. [3] Self WH1, Speroff T, Grijalva CG, McNaughton CD, Ashburn J, Liu D, et al. Reducing blood culture contamination in the emergency department: An interrupted time series quality improvement study. Acad Emerg Med. 2013;20:89–97
ED
[4] Hall RT, Domenico HJ, Self WH, Hain PD. Reducing the blood culture contamination rate in a pediatric emergency department and subsequent cost savings. Pediatrics. 2013;131:e292–297.
PT
[5] Fontanals D, Sanfeliu I, Pons I, Mariscal D, Torra M. Evaluation of the BacT/Alert and VITAL blood culture systems for the diagnosis of bacteremia. Clin Microbiol
CE
Infect. 1998 Feb;4(2):88-93.
AC
[6] Cunha BA, Infectious diseases in critical care medicine, Third Edition. 2009; 51 [7] Lee A, Mirrett S, Reller LB, Weinstein MP. Detection of bloodstream infections in adults: how many blood cultures are needed? J Clin Microbiol. 2007;45(11):3546-8. [8] Cockerill FR 3rd, Wilson JW, Vetter EA, Goodman KM, Torgerson CA, Harmsen WS, et al. Optimal testing parameters for blood cultures. Clin Infect Dis. 2004;38(12):1724-30. [9] Murray PR, Masur H. Current approaches to the diagnosis of bacterial and fungal bloodstream infections in the intensive care unit. Crit Care Med. 2012; 40(12):3277-82. [10] Pasqualini L, Mencacci A, Leli C, Montagna P, Cardaccia A, Cenci E, et al. Diagnostic performance of a multiple real-time PCR assay in patients with
ACCEPTED MANUSCRIPT
T
suspected sepsis hospitalized in an internal medicine ward. J Clin Microbiol. 2012;50(4):1285-8.
RI P
[11] Pierrakos C, Vincent JL. Sepsis biomarkers: a review. Crit Care. 2010;14(1):R15.
SC
[12] Palmiere C, Augsburger M. Markers for sepsis diagnosis in the forensic setting: state of the art. Croat Med J. 2014; 55(2):103-14. [13] Assicot M, Gendrel D, Carsin H, Raymond J, Guilbaud J, Bohuon C. High serum
MA NU
procalcitonin concentrations in patients with sepsis and infection. Lancet. 1993;27;341(8844):515-8.
ED
[14] Aikawa N, Fujishima S, Endo S, Sekine I, Kogawa K, Yamamoto Y, et al. Multicenter prospective study of procalcitonin as an indicator of sepsis. J Infect Chemother. 2005;11(3):152-9.
PT
[15] Rajkumari N, Mathur P, Sharma S, Gupta B, Bhoi S, Misra MC. Procalcitonin as a predictor of sepsis and outcome in severe trauma patients: a prospective study. J Lab Physicians. 2013;5(2):100-8.
AC
CE
[16] Nargis W, Ibrahim M, Ahamed BU. Procalcitonin versus C-reactive protein: Usefulness as biomarker of sepsis in ICU patient. Int J Crit Illn Inj Sci. 2014; 4(3):195-9. [17] Fendler WM, Piotrowski AJ. Procalcitonin in the early diagnosis of nosocomial sepsis in preterm neonates. J Paediatr Child Health. 2008; 44(3):114-8. [18] Charles PE, Ladoire S, Aho S, Quenot JP, Doise JM, Prin S, et al. Serum procalcitonin elevation in critically ill patients at the onset of bacteremia caused by either Gram negative or Gram positive bacteria. BMC Infect Dis. 2008;26(8):38. [19] Brodská H, Malíčková K, Adámková V, Benáková H, Štastná MM, Zima T. Significantly higher procalcitonin levels could differentiate Gram-negative sepsis from Gram-positive and fungal sepsis. Clin Exp Med. 2013;13(3):165-70.
ACCEPTED MANUSCRIPT
T
[20] Friend KE, Burgess JN, Britt RC, Collins JN, Weireter LN, Novosel TJ, et al. Procalcitonin elevation suggests a septic source. Am Surg. 2014;80(9):906-9.
RI P
[21] Abdi H, Williams LJ. Principal component analysis. Wiley Interdisciplinary Reviews: Computational Statistics. 2010; 2(4):433-59.
MA NU
SC
[22] Arai T, Kumasaka K, Nagata K, Okita T, Oomura T, Hoshiai A, et al. Prediction of blood culture results by measuring procalcitonin levels and other inflammatory biomarkers. Am J Emerg Med. 2014;32(4):330-3. [23] Weinstein MP. Blood culture contamination: persisting problems and partial progress. J. Clin. Microbiol. 2003;41:2275-8.
ED
[24] Roberts FJ, Geere IW, Coldman A. A three-year study of positive blood cultures, with emphasis on prognosis. Rev Infect Dis. 1991;13(1):34-46.
AC
CE
PT
[25] Guo SY, Zhou Y, Hu QF, Yao J, Wang H. Procalcitonin is a marker of gramnegative bacteremia in patients with sepsis. Am J Med Sci. 2015; 349(6):499-504.
ACCEPTED MANUSCRIPT
Table 1. Background of Patients Characteristic Male, n (%)
207 (61.8%)
T
Sex
n = 315 108 (38.2%)
RI P
Female, n (%) Age (years), mean ± SD
65.4 ± 19.5 89 (21.1%)
SC
Mortality, n (%) Title
code Number J
86
A, B
67
I
54
K
52
S, T
50
Endocrine, nutritional and metabolic diseases
E
36
Diseases of the genitourinary system
N
26
Diseases of the nervous system
G
25
Diseases of the skin and subcutaneous tissue
L
9
Diseases of the musculoskeletal system and connective tissue
M
8
C, D
4
R
3
involving the immune mechanism
D
1
Pregnancy, childbirth and the puerperium
O
1
MA NU
Diseases of the respiratory system
*ICD
Certain infectious and parasitic disease Diseases of the circulatory system Diseases of the digestive system
Neoplasms
CE
PT
ED
Injuries, poisoning and certain other consequences of external causes
Symptoms, signs and abnormal clinical and laboratory findings,
AC
not classified elsewhere
Diseases of the blood and blood-forming organs and certain disorders
*ICD = International Statistical Classification of Diseases.
ACCEPTED MANUSCRIPT
Table 2. Loading of the variables on the 2 components Component 2
0.706
0.282
PCT
0.839
0.001
PLT
- 0.644
0.579
WBC
0.200
0.868
SC
CRP
CRP, C-reactive protein
MA NU
PCT, Procalcitonin PLT, Platelet
AC
CE
PT
ED
WBC, White blood cell
T
Component 1
RI P
Variable
ACCEPTED MANUSCRIPT
Table 3. Comparisons of blood culture-positive and blood culture-negative
T
proportions classified by component 1
RI P
Component 1 -1 ≤ 0 ˂
0≤1˂
1≤2˂
2≤
Positive (n)
0
15
16
8
8
Negative (n)
21
65
35
3
0
0
19
31
73
100
100
81
69
27
0
51
11
8
Negative proportions (%)
21
AC
CE
PT
ED
Total (n)
MA NU
Positive proportions (%)
SC
< -1
80
ACCEPTED MANUSCRIPT
Table 4. Comparisons of GNR and GPC proportions classified by component 1
0≤1˂
1≤2˂
RI P
-1 ≤ 0 ˂
< -1
T
Component 1 2≤
0
2
6
5
6
GPC (n)
0
13
10
3
2
GNR proportions (%)
-
13
38
63
75
GPC proportions (%)
-
87
62
37
25
Total (n)
0
15
16
8
8
MA NU
GNR, Gram-negative rods
CE
PT
ED
GPC, Gram-positive cocci
AC
SC
GNR (n)
AC
CE
PT
ED
MA NU
SC
RI P
T
ACCEPTED MANUSCRIPT
AC
CE
PT
ED
MA NU
SC
RI P
T
ACCEPTED MANUSCRIPT