American Journal of Emergency Medicine 32 (2014) 330–333
Contents lists available at ScienceDirect
American Journal of Emergency Medicine journal homepage: www.elsevier.com/locate/ajem
Original Contribution
Prediction of blood culture results by measuring procalcitonin levels and other inflammatory biomarkers Takao Arai, MD ⁎, Kenichiro Kumasaka, MD, Katuhiro Nagata, MD, Taihei Okita, MD, Taishi Oomura, MD, Akira Hoshiai, MD, Masaharu Koyama, PhD, Shoichi Ohta, MD, PhD, Tetsuo Yukioka, MD, PhD 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
a r t i c l e
i n f o
Article history: Received 10 October 2013 Accepted 15 December 2013
a b s t r a c t Background: It would be helpful if we could predict positive or negative blood culture results. This study considered the usefulness of measuring procalcitonin (PCT) level and standard clinical biomarkers such as white blood cell (WBC) count, C-reactive protein (CRP) level, and platelet (PLT) count to predict blood culture results. Method: We retrospectively analyzed the data from 422 specimens collected at our emergency center within the preceding 36 consecutive months. Primary component analysis (PCA) was used for detecting the degree of the relational contribution of each of the 4 biomarkers to the blood culture results. Results: Procalcitonin alone (cut-off value, 0.5 ng/mL) yielded a positive blood culture rate of 34.0%. Procalcitonin plus 3 biomarkers (WBC, CRP, and PLT) analyzed by PCA yielded 45.9% or 35.3% when a case was in the first or fourth quadrant, which was significantly higher than cases in the second or third quadrant. Primary component analysis also revealed that positive blood culture results were mainly affected by primary component 1, to which PCT and PLT (not WBC or CRP) predominantly contribute. Conclusion: Although it is difficult to predict blood culture results, even using 4 biomarkers analyzed by PCA, our new finding that blood culture results are affected not by WBC and CRP, but mainly by PCT and PLT, might help explain the mechanism of sepsis. © 2014 Elsevier Inc. All rights reserved.
1. Introduction In patients with high fever and suspicion of acute infectious illness, timely and adequate clinical decision making and blood sampling are required [1]. In cases where sepsis is suspected, blood culture is recommended [1]. However, the significance of positive blood culture rate prediction is not yet fully clarified [2,3]. If we could predict the positive rate of blood cultures by quick laboratory tests in which results can be available within an hour, it will allow making any clinical decision without the result of blood culture. Procalcitonin (PCT) polypeptide, consisting of 116 amino acids, is generally synthesized by the C cells in the thyroid gland as a precursor of calcitonin [4]. In bacterial infections, however, toxins induce the production of inflammatory cytokines, which, in turn, stimulate other organs including the lungs, kidneys, and liver as well as fat cells and muscles to secrete PCT into the bloodstream [5,6]. In viral infections, PCT production is suppressed by an increase in the concentration of interferon γ, and as a result, it is conjectured that PCT concentration increases specifically in response to the bacterial infections [7]. In this study, we selected PCT and standard clinical biomarkers such as white blood cell (WBC) count, C-reactive protein (CRP), and ⁎ Corresponding author. E-mail address: [email protected] (T. Arai). 0735-6757/$ – see front matter © 2014 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.ajem.2013.12.035
platelet (PLT) count for multivariate analysis to predict blood culture results and for further clinical decision making. 2. Methods 2.1. Subjects This is a retrospective observational clinical study with approval of the ethics committee of Tokyo Medical University. At the Life Saving Emergency Center of Tokyo Medical University Hachioji Medical Center, patients with (1) fever of more than 38°C, (2) chills and shivering, or (3) suspicion of bacterial or fungal infection had blood samples taken for laboratory tests and blood culture. In a consecutive 36-month study period, 422 specimens were included in the analysis. 2.2. Measurements Data of 4 sets of laboratory tests, namely, PCT, CRP, PLT, and WBC, were analyzed. Serum concentrations of PCT were measured by chemiluminescent enzyme immunoassay using Sphere Light 180 (Beckman Coulter, Inc, Brea, CA), whereas serum concentrations of CRP were measured by luminescent oxygen channeling immunoassay using Dimension Vista 1500 T (Siemens, Munich, Germany). White blood cell and PLT counts were performed using the Cell-Dyn Sapphire
T. Arai et al. / American Journal of Emergency Medicine 32 (2014) 330–333
(Abbott Diagnostics Division, Santa Clara, CA). Blood culture results were obtained using the BACTEC 9120 (Japan Becton, Dickinson and Company, Tokyo, Japan). If only 1 set of coagulase-negative staphylococci was detected, we considered it to be contamination and judged the result as “negative.” 2.3. Statistical analysis Statistical analyses were performed using SPSS version 19 (SPSS, Chicago, IL). Differences in the mean concentrations of PCT, CRP, PLT, and WBC between 2 groups were analyzed using the t test. Comparisons between proportions were made using the χ2 test. P b .05 was considered to indicate a statistically significant difference. To investigate the contribution of each test, we standardized the data and performed principal component analysis (PCA). Principal component analysis is a multivariate technique that analyzes data tables in which observations are described by several intercorrelated quantitative-dependent variables. Its goal is to extract important information from the table, represent it as a set of new orthogonal variables called principal components, and display the pattern of similarities of the observations and variables as points in maps [8]. To choose the components, we fixed the number of components to 2. 3. Results The mean age, sex, mortality, and principal diagnosis (International Statistical Classification of Diseases, 10th Revision) are shown in Table 1, and the mean values of the laboratory tests are shown in Table 2. In patients with positive blood cultures, the PCT and CRP values were significantly higher than in patients with negative blood cultures. Platelet was significantly lower in the positive blood culture group. There was no significant difference in the WBC count between the 2 groups. Recently, a cut-off value of 0.5 ng/mL of blood PCT has been commonly used for the diagnosis of sepsis [4]. As a result of this cutoff value, patients with positive PCT had significantly higher positive Table 1 Patients' background Numbers or average ± SD (n = 422) Age (y) Sex
Male (%) Female (%)
67.86 ± 16.496 261 (61.8%) 161 (38.2%) 89 (21.1%)
Title
ICD code
n
Diseases of the respiratory system Symptoms, signs, and abnormal clinical and laboratory findings, not classified elsewhere Certain infectious and parasitic disease Diseases of the digestive system Injuries, poisoning, and certain other consequences of external causes Diseases of the circulatory system Endocrine, nutritional, and metabolic diseases Diseases of the genitourinary system Diseases of the skin and subcutaneous tissue Neoplasms Diseases of the nervous system Diseases of the musculoskeletal system and connective tissue Diseases of the blood and blood-forming organs and certain disorders Involving the immune mechanism Pregnancy, childbirth, and the puerperium
J R
86 3
A, B K S, T
67 52 50
I E
54 36
N L
26 9
C, D G M
4 25 8
Mortality (%)
D O
331
Table 2 Descriptive statistics (a): Descriptive statistics of 4 biomarkers
Minimum Maximum Mean SD
CRP (mg/dL)
PCT (ng/μL)
PLT (104/μL)
WBC (/μL)
0.01 41.43 11.90 9.20
b0.01 200.0 15.91 36.25
0.80 139.5 21.68 14.83
641 59750 13433.3 8213.4
(b): Comparison of mean values of 4 biomarkers in positive and negative blood culture groups Mean Positive CRP (mg/dL) PCT (ng/μL) PLT (104/μL) WBC (/μL)
14.11 30.62 19.01 14438.9
P Negative 11.08 10.47 22.68 13061.1
0.003* b0.001* 0.024* 0.175
* Statistically significant difference.
blood culture rates than patients with negative PCT (P b .05, χ 2 test, Table 3). However, positive or negative PCT rates are not sufficient to allow the use PCT alone as an index to predict blood culture results. The results of PCA using WBC, PLT, CRP, and PCT showed that the proportion of variance of principal component 1 was 32% and that of 2 was 29%. That is, a total of 61% of information of the 4 parameters (4-dimensional data) was compressed in 2 dimensional, which comprised principal components 1 and 2. Fig. 1 shows the factor loading of PCT, WBC, PLT, and CRP contributing to principal components 1 and 2. In individual cases, the individual scores of principal components 1 and 2 are calculated. On the score calculation of principal component 1, PCT contributes as a plus factor, whereas PLT is a minus factor, and WBC hardly affects the score calculation because of its small coefficient. In other words, the score of principal component 1 rises when PCT is high and PLT is low. The degree of influence of WBC is only 1/8 compared with PCT. On the other hand, with principal component 2, the effect of WBC is the largest, followed by CRP, which is similar to PLT. The contribution of PCT is as small as 1/5 that of WBC. Fig. 2 shows the distribution of individual clinical cases identifying the results of the blood culture. The horizontal axis represents principal component 1, and the vertical axis represents principal component 2. On the first, second, third, and fourth quadrants, there are 98, 114, 130, and 85 patients, respectively. Table 4 shows the proportion of the positive blood culture result of each quadrant. The proportion of the positive blood culture was significantly higher in the first and fourth quadrants than in the second and third quadrants (P b .05, χ2 test, Table 5). 4. Discussion The results showed that PCT, a new laboratory test for bacterial infection, alone cannot be used as an index for the prediction of blood culture results. Any other single popular laboratory test may not be a good predictor. This is the reason why we applied PCA including PCT with 3 other common laboratory tests. Primary component analysis is one of the multivariate analyses focusing on data variance and is often used to compress multidimensional data into 2-dimensional visualized information [8]. In our study, 61% of the total variance of 4-dimensional data composed by PCT, CRP, PLT, and WBC was captured into 2 new dimensions or components. Table 3 Comparison of positive blood culture rates and the cut-off value Cut-off value (ng/mL)
Positive blood culture rate Less than cut-off value
Greater than cut-off value
0.5
13.7% (20/146)
34.0 (94/276)
1 1
P
.05
332
T. Arai et al. / American Journal of Emergency Medicine 32 (2014) 330–333 Table 4 Proportion of positive blood culture result of each quadrant
Fig. 1. Factor loading of PCT, WBC, PLT, and CRP contributing to principal components 1 and 2. On score calculation of principal component 1, PCT contributes as a plus factor, whereas PLT is minus, and WBC hardly affects score calculation because of its small coefficient. That is, principal component 1 presents the bacterial (fungal) infection and endothelial injury. For principal component 2, the effect of WBC is the largest, followed by CRP, which is similar to PLT. The contribution of PCT is as small as 1/5 that of WBC. Principal component 2 presents systemic inflammation.
On principal component 1, the factor loading of PCT was 0.82 and that of PLT was − 0.65 (Fig. 1). That is, when PCT is high and PLT is low, the score of principal component 1 would be high. The influence of WBC on principal component 1 is by far lower than that of PCT and PLT. That is, the principal component 1 score increases with bacterial or fungal infection, possibly with vascular endothelial injuries, which lead to consumption coagulopathy and reduction in PLT count. Principal component 1 is regarded as a new synthesized parameter related to a direct reaction to bacterial infection. On principal component 2, WBC, the most common marker of systemic inflammation, had the largest influence with the factor loading at 0.79. The factor loadings of CRP and PLT were nearly equal
Quadrant
Component 1
Component 2
Blood culture Positive
Negative
1 2 3 4
+ + − −
+ − − +
45 (45.90%) 19 (16.70%) 20 (15.40%) 30 (35.30%)
53 (54.10%) 95 (83.30%) 110 (84.60%) 55 (64.70%)
at approximately 0.5. The factor loading of PCT was low at 0.2. Therefore, principal component 2 represented the systemic inflammation response, not the index of the reaction of specific bacterial or fungal infection. When principal components 1 and 2 were positive and a case was in the first quadrant, it indicated that not only specific reaction to infection but also systemic inflammatory response was prominent in the patient. In this circumstance, it was expected that the blood culture is positive. On the other hand, when principal components 1 and 2 were negative and a case was in the third quadrant, it was expected that the blood culture was negative. This is because of a weak direct reaction to bacterial infection. The results of our study correspond to this assumption. Nearly half of the patients (46%) in the first quadrant had positive blood cultures, whereas 15% or only 1/3 of that of the first quadrant had positive blood cultures in the third quadrant. Patients in the fourth quadrant (principal component 1, positive; principal component 2, negative) had positive blood culture rates close to those in the first quadrant (35%), and patients in the second quadrant (principal component 1, negative; principal component 2, positive) had positive blood culture results close to those in the third quadrant (17%). These results suggest that principal component 1 can be used to predict blood culture results, and this corresponds with our assumption. Our results suggest that, with 4 common laboratory tests in the emergency department, a patient's condition in Fig. 2 can be identified, and when a patient is in the first or fourth quadrant, blood cultures should be performed immediately to identify the organisms. For patients in the second or third quadrant, blood culture should require less immediate attention. We need to add that, in each case, it was difficult to predict blood culture results with 4 laboratory tests, even with PCA. As for our novel finding, we found that blood culture results are not affected by WBC and CRP, but mainly by PCT and PLT. This suggests that just as with bacterial or fungal infection, the presence of vascular endothelial injuries may also affect blood culture results. This finding might suggest possible associations to further clarify the “mechanism of sepsis,” which warrants more detailed examination. In addition, further prospective studies will be required to identify patients in whom blood culture is essential for clinical decision making. Acknowledgments The authors thank Ms Maya Vardaman and Associate Prof Edward F. Barroga (PhD) of the Department of International Medical Communications of Tokyo Medical University for their editorial review of the English manuscript. Table 5 Comparison of blood culture positive rates between the first and fourth quadrants and the second and third quadrants Quadrant
Fig. 2. Distribution of individual clinical cases identifying the results of blood culture. The proportion of positive blood culture rates is significantly higher in the first and fourth quadrants than in the second and third quadrants, as shown in Table 5.
1 and 4 2 and 3 χ2 Value: 33.396; P b .01.
Blood culture
Total
Positive
Negative
75 (41.0%) 39 (16.0%) 114
108 (59.0%) 205 (84.0%) 313
183 244 427
T. Arai et al. / American Journal of Emergency Medicine 32 (2014) 330–333
References [1] Dellinger RP, et al. Surviving sepsis campaign: international guidelines for management of severe sepsis and septic shock: 2012. Crit Care Med 2013;41(2): 580–637. [2] Groeneveld AB, et al. Circulating inflammatory mediators in patients with fever: predicting bloodstream infection. Clin Diagn Lab Immunol 2001;8(6):1189–95. [3] Chirouze C, Schumacher H, Rabaud C. Low serum procalcitonin level accurately predicts the absence of bacteremia in adult patients with acute fever. Clin Infect Dis 2002;35(2):156–61.
333
[4] Christ-Crain M, Müller B. Procalcitonin in bacterial infections—hype, hope, more or less? Swiss Med Wkly 2005;135(31–32):451–60. [5] Müller B, Becker KL, Schächinger H. Calcitonin precursors are reliable markers of sepsis in a medical intensive care unit. Crit Care Med 2000;28(4):977–83. [6] Müller B, et al. Ubiquitous expression of the calcitonin-i gene in multiple tissues in response to sepsis. J Clin Endocrinol Metab 2001;86(1):396–404. [7] Linscheid P, Seboek D, Nylon ES. In vitro and in vivo calcitonin I gene expression in parenchymal cells: a novel product of human adipose tissue. Endocrinology 2003;144(12):5578–84. [8] Abdi H, Williams LJ. Principal component analysis, 2. John Wiley & Sons, Inc. WIREs Comp Stat; 2010 433–59.
Contents lists available at ScienceDirect
American Journal of Emergency Medicine journal homepage: www.elsevier.com/locate/ajem
Original Contribution
Prediction of blood culture results by measuring procalcitonin levels and other inflammatory biomarkers Takao Arai, MD ⁎, Kenichiro Kumasaka, MD, Katuhiro Nagata, MD, Taihei Okita, MD, Taishi Oomura, MD, Akira Hoshiai, MD, Masaharu Koyama, PhD, Shoichi Ohta, MD, PhD, Tetsuo Yukioka, MD, PhD 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
a r t i c l e
i n f o
Article history: Received 10 October 2013 Accepted 15 December 2013
a b s t r a c t Background: It would be helpful if we could predict positive or negative blood culture results. This study considered the usefulness of measuring procalcitonin (PCT) level and standard clinical biomarkers such as white blood cell (WBC) count, C-reactive protein (CRP) level, and platelet (PLT) count to predict blood culture results. Method: We retrospectively analyzed the data from 422 specimens collected at our emergency center within the preceding 36 consecutive months. Primary component analysis (PCA) was used for detecting the degree of the relational contribution of each of the 4 biomarkers to the blood culture results. Results: Procalcitonin alone (cut-off value, 0.5 ng/mL) yielded a positive blood culture rate of 34.0%. Procalcitonin plus 3 biomarkers (WBC, CRP, and PLT) analyzed by PCA yielded 45.9% or 35.3% when a case was in the first or fourth quadrant, which was significantly higher than cases in the second or third quadrant. Primary component analysis also revealed that positive blood culture results were mainly affected by primary component 1, to which PCT and PLT (not WBC or CRP) predominantly contribute. Conclusion: Although it is difficult to predict blood culture results, even using 4 biomarkers analyzed by PCA, our new finding that blood culture results are affected not by WBC and CRP, but mainly by PCT and PLT, might help explain the mechanism of sepsis. © 2014 Elsevier Inc. All rights reserved.
1. Introduction In patients with high fever and suspicion of acute infectious illness, timely and adequate clinical decision making and blood sampling are required [1]. In cases where sepsis is suspected, blood culture is recommended [1]. However, the significance of positive blood culture rate prediction is not yet fully clarified [2,3]. If we could predict the positive rate of blood cultures by quick laboratory tests in which results can be available within an hour, it will allow making any clinical decision without the result of blood culture. Procalcitonin (PCT) polypeptide, consisting of 116 amino acids, is generally synthesized by the C cells in the thyroid gland as a precursor of calcitonin [4]. In bacterial infections, however, toxins induce the production of inflammatory cytokines, which, in turn, stimulate other organs including the lungs, kidneys, and liver as well as fat cells and muscles to secrete PCT into the bloodstream [5,6]. In viral infections, PCT production is suppressed by an increase in the concentration of interferon γ, and as a result, it is conjectured that PCT concentration increases specifically in response to the bacterial infections [7]. In this study, we selected PCT and standard clinical biomarkers such as white blood cell (WBC) count, C-reactive protein (CRP), and ⁎ Corresponding author. E-mail address: [email protected] (T. Arai). 0735-6757/$ – see front matter © 2014 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.ajem.2013.12.035
platelet (PLT) count for multivariate analysis to predict blood culture results and for further clinical decision making. 2. Methods 2.1. Subjects This is a retrospective observational clinical study with approval of the ethics committee of Tokyo Medical University. At the Life Saving Emergency Center of Tokyo Medical University Hachioji Medical Center, patients with (1) fever of more than 38°C, (2) chills and shivering, or (3) suspicion of bacterial or fungal infection had blood samples taken for laboratory tests and blood culture. In a consecutive 36-month study period, 422 specimens were included in the analysis. 2.2. Measurements Data of 4 sets of laboratory tests, namely, PCT, CRP, PLT, and WBC, were analyzed. Serum concentrations of PCT were measured by chemiluminescent enzyme immunoassay using Sphere Light 180 (Beckman Coulter, Inc, Brea, CA), whereas serum concentrations of CRP were measured by luminescent oxygen channeling immunoassay using Dimension Vista 1500 T (Siemens, Munich, Germany). White blood cell and PLT counts were performed using the Cell-Dyn Sapphire
T. Arai et al. / American Journal of Emergency Medicine 32 (2014) 330–333
(Abbott Diagnostics Division, Santa Clara, CA). Blood culture results were obtained using the BACTEC 9120 (Japan Becton, Dickinson and Company, Tokyo, Japan). If only 1 set of coagulase-negative staphylococci was detected, we considered it to be contamination and judged the result as “negative.” 2.3. Statistical analysis Statistical analyses were performed using SPSS version 19 (SPSS, Chicago, IL). Differences in the mean concentrations of PCT, CRP, PLT, and WBC between 2 groups were analyzed using the t test. Comparisons between proportions were made using the χ2 test. P b .05 was considered to indicate a statistically significant difference. To investigate the contribution of each test, we standardized the data and performed principal component analysis (PCA). Principal component analysis is a multivariate technique that analyzes data tables in which observations are described by several intercorrelated quantitative-dependent variables. Its goal is to extract important information from the table, represent it as a set of new orthogonal variables called principal components, and display the pattern of similarities of the observations and variables as points in maps [8]. To choose the components, we fixed the number of components to 2. 3. Results The mean age, sex, mortality, and principal diagnosis (International Statistical Classification of Diseases, 10th Revision) are shown in Table 1, and the mean values of the laboratory tests are shown in Table 2. In patients with positive blood cultures, the PCT and CRP values were significantly higher than in patients with negative blood cultures. Platelet was significantly lower in the positive blood culture group. There was no significant difference in the WBC count between the 2 groups. Recently, a cut-off value of 0.5 ng/mL of blood PCT has been commonly used for the diagnosis of sepsis [4]. As a result of this cutoff value, patients with positive PCT had significantly higher positive Table 1 Patients' background Numbers or average ± SD (n = 422) Age (y) Sex
Male (%) Female (%)
67.86 ± 16.496 261 (61.8%) 161 (38.2%) 89 (21.1%)
Title
ICD code
n
Diseases of the respiratory system Symptoms, signs, and abnormal clinical and laboratory findings, not classified elsewhere Certain infectious and parasitic disease Diseases of the digestive system Injuries, poisoning, and certain other consequences of external causes Diseases of the circulatory system Endocrine, nutritional, and metabolic diseases Diseases of the genitourinary system Diseases of the skin and subcutaneous tissue Neoplasms Diseases of the nervous system Diseases of the musculoskeletal system and connective tissue Diseases of the blood and blood-forming organs and certain disorders Involving the immune mechanism Pregnancy, childbirth, and the puerperium
J R
86 3
A, B K S, T
67 52 50
I E
54 36
N L
26 9
C, D G M
4 25 8
Mortality (%)
D O
331
Table 2 Descriptive statistics (a): Descriptive statistics of 4 biomarkers
Minimum Maximum Mean SD
CRP (mg/dL)
PCT (ng/μL)
PLT (104/μL)
WBC (/μL)
0.01 41.43 11.90 9.20
b0.01 200.0 15.91 36.25
0.80 139.5 21.68 14.83
641 59750 13433.3 8213.4
(b): Comparison of mean values of 4 biomarkers in positive and negative blood culture groups Mean Positive CRP (mg/dL) PCT (ng/μL) PLT (104/μL) WBC (/μL)
14.11 30.62 19.01 14438.9
P Negative 11.08 10.47 22.68 13061.1
0.003* b0.001* 0.024* 0.175
* Statistically significant difference.
blood culture rates than patients with negative PCT (P b .05, χ 2 test, Table 3). However, positive or negative PCT rates are not sufficient to allow the use PCT alone as an index to predict blood culture results. The results of PCA using WBC, PLT, CRP, and PCT showed that the proportion of variance of principal component 1 was 32% and that of 2 was 29%. That is, a total of 61% of information of the 4 parameters (4-dimensional data) was compressed in 2 dimensional, which comprised principal components 1 and 2. Fig. 1 shows the factor loading of PCT, WBC, PLT, and CRP contributing to principal components 1 and 2. In individual cases, the individual scores of principal components 1 and 2 are calculated. On the score calculation of principal component 1, PCT contributes as a plus factor, whereas PLT is a minus factor, and WBC hardly affects the score calculation because of its small coefficient. In other words, the score of principal component 1 rises when PCT is high and PLT is low. The degree of influence of WBC is only 1/8 compared with PCT. On the other hand, with principal component 2, the effect of WBC is the largest, followed by CRP, which is similar to PLT. The contribution of PCT is as small as 1/5 that of WBC. Fig. 2 shows the distribution of individual clinical cases identifying the results of the blood culture. The horizontal axis represents principal component 1, and the vertical axis represents principal component 2. On the first, second, third, and fourth quadrants, there are 98, 114, 130, and 85 patients, respectively. Table 4 shows the proportion of the positive blood culture result of each quadrant. The proportion of the positive blood culture was significantly higher in the first and fourth quadrants than in the second and third quadrants (P b .05, χ2 test, Table 5). 4. Discussion The results showed that PCT, a new laboratory test for bacterial infection, alone cannot be used as an index for the prediction of blood culture results. Any other single popular laboratory test may not be a good predictor. This is the reason why we applied PCA including PCT with 3 other common laboratory tests. Primary component analysis is one of the multivariate analyses focusing on data variance and is often used to compress multidimensional data into 2-dimensional visualized information [8]. In our study, 61% of the total variance of 4-dimensional data composed by PCT, CRP, PLT, and WBC was captured into 2 new dimensions or components. Table 3 Comparison of positive blood culture rates and the cut-off value Cut-off value (ng/mL)
Positive blood culture rate Less than cut-off value
Greater than cut-off value
0.5
13.7% (20/146)
34.0 (94/276)
1 1
P
.05
332
T. Arai et al. / American Journal of Emergency Medicine 32 (2014) 330–333 Table 4 Proportion of positive blood culture result of each quadrant
Fig. 1. Factor loading of PCT, WBC, PLT, and CRP contributing to principal components 1 and 2. On score calculation of principal component 1, PCT contributes as a plus factor, whereas PLT is minus, and WBC hardly affects score calculation because of its small coefficient. That is, principal component 1 presents the bacterial (fungal) infection and endothelial injury. For principal component 2, the effect of WBC is the largest, followed by CRP, which is similar to PLT. The contribution of PCT is as small as 1/5 that of WBC. Principal component 2 presents systemic inflammation.
On principal component 1, the factor loading of PCT was 0.82 and that of PLT was − 0.65 (Fig. 1). That is, when PCT is high and PLT is low, the score of principal component 1 would be high. The influence of WBC on principal component 1 is by far lower than that of PCT and PLT. That is, the principal component 1 score increases with bacterial or fungal infection, possibly with vascular endothelial injuries, which lead to consumption coagulopathy and reduction in PLT count. Principal component 1 is regarded as a new synthesized parameter related to a direct reaction to bacterial infection. On principal component 2, WBC, the most common marker of systemic inflammation, had the largest influence with the factor loading at 0.79. The factor loadings of CRP and PLT were nearly equal
Quadrant
Component 1
Component 2
Blood culture Positive
Negative
1 2 3 4
+ + − −
+ − − +
45 (45.90%) 19 (16.70%) 20 (15.40%) 30 (35.30%)
53 (54.10%) 95 (83.30%) 110 (84.60%) 55 (64.70%)
at approximately 0.5. The factor loading of PCT was low at 0.2. Therefore, principal component 2 represented the systemic inflammation response, not the index of the reaction of specific bacterial or fungal infection. When principal components 1 and 2 were positive and a case was in the first quadrant, it indicated that not only specific reaction to infection but also systemic inflammatory response was prominent in the patient. In this circumstance, it was expected that the blood culture is positive. On the other hand, when principal components 1 and 2 were negative and a case was in the third quadrant, it was expected that the blood culture was negative. This is because of a weak direct reaction to bacterial infection. The results of our study correspond to this assumption. Nearly half of the patients (46%) in the first quadrant had positive blood cultures, whereas 15% or only 1/3 of that of the first quadrant had positive blood cultures in the third quadrant. Patients in the fourth quadrant (principal component 1, positive; principal component 2, negative) had positive blood culture rates close to those in the first quadrant (35%), and patients in the second quadrant (principal component 1, negative; principal component 2, positive) had positive blood culture results close to those in the third quadrant (17%). These results suggest that principal component 1 can be used to predict blood culture results, and this corresponds with our assumption. Our results suggest that, with 4 common laboratory tests in the emergency department, a patient's condition in Fig. 2 can be identified, and when a patient is in the first or fourth quadrant, blood cultures should be performed immediately to identify the organisms. For patients in the second or third quadrant, blood culture should require less immediate attention. We need to add that, in each case, it was difficult to predict blood culture results with 4 laboratory tests, even with PCA. As for our novel finding, we found that blood culture results are not affected by WBC and CRP, but mainly by PCT and PLT. This suggests that just as with bacterial or fungal infection, the presence of vascular endothelial injuries may also affect blood culture results. This finding might suggest possible associations to further clarify the “mechanism of sepsis,” which warrants more detailed examination. In addition, further prospective studies will be required to identify patients in whom blood culture is essential for clinical decision making. Acknowledgments The authors thank Ms Maya Vardaman and Associate Prof Edward F. Barroga (PhD) of the Department of International Medical Communications of Tokyo Medical University for their editorial review of the English manuscript. Table 5 Comparison of blood culture positive rates between the first and fourth quadrants and the second and third quadrants Quadrant
Fig. 2. Distribution of individual clinical cases identifying the results of blood culture. The proportion of positive blood culture rates is significantly higher in the first and fourth quadrants than in the second and third quadrants, as shown in Table 5.
1 and 4 2 and 3 χ2 Value: 33.396; P b .01.
Blood culture
Total
Positive
Negative
75 (41.0%) 39 (16.0%) 114
108 (59.0%) 205 (84.0%) 313
183 244 427
T. Arai et al. / American Journal of Emergency Medicine 32 (2014) 330–333
References [1] Dellinger RP, et al. Surviving sepsis campaign: international guidelines for management of severe sepsis and septic shock: 2012. Crit Care Med 2013;41(2): 580–637. [2] Groeneveld AB, et al. Circulating inflammatory mediators in patients with fever: predicting bloodstream infection. Clin Diagn Lab Immunol 2001;8(6):1189–95. [3] Chirouze C, Schumacher H, Rabaud C. Low serum procalcitonin level accurately predicts the absence of bacteremia in adult patients with acute fever. Clin Infect Dis 2002;35(2):156–61.
333
[4] Christ-Crain M, Müller B. Procalcitonin in bacterial infections—hype, hope, more or less? Swiss Med Wkly 2005;135(31–32):451–60. [5] Müller B, Becker KL, Schächinger H. Calcitonin precursors are reliable markers of sepsis in a medical intensive care unit. Crit Care Med 2000;28(4):977–83. [6] Müller B, et al. Ubiquitous expression of the calcitonin-i gene in multiple tissues in response to sepsis. J Clin Endocrinol Metab 2001;86(1):396–404. [7] Linscheid P, Seboek D, Nylon ES. In vitro and in vivo calcitonin I gene expression in parenchymal cells: a novel product of human adipose tissue. Endocrinology 2003;144(12):5578–84. [8] Abdi H, Williams LJ. Principal component analysis, 2. John Wiley & Sons, Inc. WIREs Comp Stat; 2010 433–59.
