Cytokine markers as predictors of type of respiratory infection in patients during the influenza season?

John Patrick Haran MD ^a,b,?, Rachel Buglione-Corbett ^c, Shan Lu MD, PhD ^c

^a Department of Emergency Medicine, University of Massachusetts Medical School, Worcester, MA 01655, USA

^b Department of Emergency Medicine, Alpert Medical School of Brown University, Providence, RI 02903, USA

^c Department of Medicine, Laboratory of Nucleic Acid Vaccines, University of Massachusetts Medical School, Worcester, MA 01655, USA

a r t i c l e i n f o

Article history:

Received 27 January 2013

Received in revised form 30 January 2013 Accepted 31 January 2013

a b s t r a c t

Objective: The objective of this study is to characterize the Cytokine response among patients presenting with an influenza-like illness who are infected with the influenza virus, a Bacterial pneumonia, or another viral infection. We hypothesize that there are differences in proinflammatory and anti-inflammatory cytokines in relation to cytokines associated with the humoral response during viral and bacterial Respiratory infections. Methods: We enrolled adults who presented to an urban academic emergency department during the 2008 to 2011 Influenza seasons with symptoms of fever and a cough. Subjects had nasal aspirates tested by viral culture, and peripheral Blood drawn to quantify cytokine concentrations. Cytokine concentrations were compared between groups using the Wilcoxon rank sum test, and receiver operating characteristic curves were calculated.

Results: A total of 80 patients were enrolled: 40 with Influenza infection, 14 patients with a bacterial pneumonia as determined by infiltrate on chest x-ray, and 26 patients negative for influenza infection and infiltrate. There were differences between the bacterial pneumonia group, and all other viral infections grouped together with regard to interleukin (IL) 4 (2.66 vs 16.77 pg/mL, P b .001), IL-5 (20.57 vs 57.57 pg/mL, P = .006), IL-6 (403.06 vs 52.69 pg/mL, P b .001), granulocyte macrophage colony-stimulating factor (18.26 vs 66.80 pg/mL, P b .001), and interferon ? (0.0 vs 830.36 pg/mL, P b .001). Interleukin 10 concentrations were elevated in patients with influenza (88.69 pg/mL) compared with all other groups combined (39.19 pg/mL; P = .003).

Conclusion: Cytokines IL-4, IL-5, IL-6, granulocyte macrophage colony-stimulating factor, and interferon ? may serve as distinct markers of bacterial infection in patients with an influenza-like illness, whereas IL-10 is uniquely elevated in influenza patients.

(C) 2013

Introduction

Currently, there is no standardized objective method to distinguish bacterial pneumonia from influenza infection or from any other viral Respiratory tract infections. Physicians rely heavily on the presence of an infiltrate on Chest x-ray to help make the diagnosis of bacterial pneumonia. Upwards of a third of patients with community- acquired pneumonia (CAP) have a normal CXR [1,2]. white blood cell count has, therefore, been used to aid in distinguishing

? Funding: This study was designed and carried out at Rhode Island Hospital/Brown University. Patient recruitment, viral testing, and sample storage were supported by an intradepartmental grant through the Department of Emergency Medicine. The cytokine

assay was supported through the University of Massachusetts Medical School’s Healey Endowment Grant.

* Corresponding author. Department of Emergency Medicine, University of Massachusetts, Worcester, MA 01655, USA.

E-mail address: [email protected] (J.P. Haran).

bacterial infection from viral infections; however, it lacks both the sensitivity and specificity to be a reliable diagnostic tool [3,4]. Ultimately, the diagnosis of bacterial infection depends on the clinician using information pooled from patient history, physical examination, and laboratory and radiological data.

During the acute phase, response to infection C-reactive protein (CRP) has been shown to be markedly elevated in bacterial infections and is associated with less of a vigorous response in viral infections [5-7]. Hepatic production of CRP varies according to the type of infection, largely in response to the cytokines that stimulate its production. The major players in this signaling of CRP production are interleukin (IL) 6, IL-1?, and tumor necrosis factor (TNF) ? [8]. Lower concentrations of CRP are observed in viral and mycoplasma respiratory infections compared with other bacterial etiologies [9]. We have previously reported on the utility of CRP in aiding emergency department (ED) physicians in the diagnosis of bacterial infection among patients with Acute respiratory infections during the influenza season [7].

In addition to CRP, cytokine concentrations have also been shown to vary according to the presence of viral vs bacterial infections.

0735-6757/$ – see front matter (C) 2013 http://dx.doi.org/10.1016/j.ajem.2013.01.030

laboratory assays“>Several studies involving pediatric patients have observed a correla- tion between sera cytokine concentrations and the Severity of disease in relation to influenza virus-associated encephalopathy [10]. A study among 19 healthy adult volunteers experimentally exposed to influenza A virus showed increasing sera cytokine concentrations that correlated with illness severity [11].

In April of 2009, a novel influenza virus emerged that was antigenically unrelated to human seasonal influenza viruses, yet related to viruses know to circulate in swine [12]. In contrast to seasonal influenza, the most serious illnesses were clustered among younger adults, with approximately 90% of the deaths occurring in patients between 18 and 65 years of age [13]. Complications from pandemic influenza infection have been associated with marked increases in certain cytokines, whereas others were significantly decreased. This resulting “cytokine storm” injures lung tissue and supports bacterial infection [14]. Pandemic H1N1 and seasonal influenza have been shown to have similar concentrations of IL-1?, IL-10, and IL-12, whereas elevated IL-8 serum concentrations were noted only in pandemic patients [15]. It has yet to be elucidated whether there are any distinct sera cytokines that may correlate with lung injury or bacterial infection or distinguish influenza infection from other viral respiratory infections. We also do not know what role, if any, specific cytokine concentrations play in patients presenting to the ED with seasonal influenza infection who are suffering from secondary bacterial infection or acute lung injury requiring hospitalization.

In light of the previously described evidence, sera cytokine

concentrations may be important and readily measurable markers of disease type and severity among adults with respiratory complaints who present to the ED during the influenza season for medical care.

Methods

Study design and setting

This was a prospective, observational investigation concerning the measurement of cytokine concentrations in the blood of adults presenting to the ED with an Influenza-Like Illness . The study setting was an urban academic level 1 trauma center with an annual census greater than 100 000 patients. Patient enrollment was conducted over 3 influenza seasons; during this period, there was the emergence of the pandemic H1N1 influenza virus. Subjects with both seasonal and H1N1 influenza infection were enrolled. The first influenza season was prepandemic and spanned the months of January through March of 2009. This subject population was initially used to demonstrate the value of CRP as a sensitive and specific marker for diagnosing bacterial infection in patients presenting with ILI during the influenza season [7].

Selection of participants

The ED records of patients of 18 years or older presenting to the adult ED were screened for chief complaints that included ILI symptoms, specifically patient reported complaints of fever and a cough. Patients were screened for enrollment during different time blocks chosen randomly from the hours of 7 AM through 1 AM 7 days a week by trained research assistants Screening patients with respira- tory complaints. These patients were approached for enrollment. Our exclusion criteria were as follows: patients presenting with symptoms lasting longer than 5 days, a history of liver disease, current regimen of immunosuppressant medications, prior diagnosis of pneumonia, or current regimen of antibiotics. In the ED, patients were surveyed about their medical history, Symptom severity, and presenting illness characteristics. A severity of symptom score was calculated for each patient in which each patient was asked to rate the severity of seven

symptoms: cough, nasal obstruction, sore throat, fatigue, headache, myalgia, and feverishness on a scale from 0, being none, to 3, being severe [16]. We collected all information using a standardized case reporting form. We screened patients over 3 influenza seasons, October through March in the years of 2008 through 2011.

After enrollment, patients were followed up by both a chart review, if they were admitted to the hospital, and with a 4-week telephone follow-up survey. The purpose of the follow-up was to determine duration of illness and any potential complications such as subsequent hospitalization and diagnosis of pneumonia. Patients were excluded from analysis if we were unable to contact them by telephone following discharge from the ED within the 4-week period. Patients were also excluded if they were diagnosed with another bacterial infectious process following their initial ED visit (ie, streptococcal pharyngitis).

Sample collection and storage

During patient enrollment in the ED, both nasal aspirates and blood samples were collected from each patient by the research team. For the nasal aspirates, the patient was placed in a seated position with the head tilted backwards gently. While occluding the opposite nostril, 5 mL of sterile saline will be inserted into the nasopharynx, held there for 5 to 10 seconds, and then gently expelled by the patient into a sterile Petri dish. This process was repeated for the opposite nostril. Samples were frozen at -80?C for storage.

Blood samples were collected by venipuncture performed by the nurse caring for the patient. Two K2E EDTA K2 vacuette tubes were collected, for a total of 8 mL of peripheral blood. The collected blood was centrifuged at 2000 to 3000 relative centrifugal force for 15 minutes at 4?C. The plasma was then immediately transferred into 1-mL aliquots and stored at -80?C.

Laboratory assays

Nasal washes were sent to the Rhode Island Hospital Virology Laboratory, where viral cultures were performed. In brief, after centrifugation of these specimens, they were processed for culture in rhesus monkey kidney cells for 2 days. Afterwards, they were assessed for the detection of viral particles using monoclonal antibodies.

Cytokine concentrations from frozen stored patient samples were analyzed using a custom Bio-Plex cytokine (Bio-Rad, Hercules, CA) multiplex bead array assay, a technology that uses digital signal processing capable of classifying polystyrene beads (microspheres) dyed with distinct proportions of red and near-infrared fluorophores. These proportions define “spectral addresses” for each bead population. As a result, up to 100 different detection reactions can be carried out simultaneously. These plates were then analyzed using flow cytometry on a Bio-Plex 200 system and Bio-Plex Manager software (Bio-Rad) [17,18]. In this study, we used the Human Cytokine Magnetic 10-plex (Bio-Rad), which includes the following cytokines: IL-1?, IL-2, IL-4, IL-5, IL-6, IL-8, IL-10, interferon ? (IFN?), TNF?, and granulocyte macrophage colony-stimulating factor (GM-CSF).

Outcomes

Patients were classified into 1 of the following 4 groups: (1) bacterial pneumonia, (2) seasonal influenza infection, (3) novel H1N1 influenza infection, and (4) other respiratory infection. The first group consisted of bacterially infected patients without concomitant influenza infection. Classification of bacterial pneumonia was deter- mined by infiltration seen on CXR at presentation, along with the diagnosis of pneumonia made by the treating physician. To determine the etiology of the viral infectious agent, we used the viral cultures to verify whether the patient was infected with either seasonal influenza

Table 1

Study population demographics divided into bacterial pneumonia, seasonal influenza, H1N1 influenza, and other infection groups

n	Bacterial		Viral all		Seasonal influenza		H1N1 influenza		Other infection	P
	14		66		23		17		26
Age (y)	57.5 (42.8-72.2)		30.4 (26.5-45.3)		30.7 (25.5-36.0)		30.8 (22.2-39.3)		29.9 (22.5-37.9)	b.001
Female	85.7% (61.1-100)		59.7% (45.9-73.5)		56.5% (31.1-81.9)		58.8% (28.4-89.2)		63.0% (40.4-85.5)	.431
White	64.3% (30.6-97.7)		46.8% (32.1-61.5)		36.8% (9.0-64.6)		35.3% (5.8-64.8)		61.5% (38.3-84.7)	.248
Flu vaccine	57.1% (22.4-91.9)		18.2% (7.2-29.2)		21.7% (0.6-42.9)		5.8% (0.0-20.4)		22.2% (2.8-41.6)	.008
Days PTA	4.1 (2.9-5.3)		2.4 (1.8-2.9)		2.6 (2.3-2.9)		1.7 (1.0-2.4)		2.6 (1.3-3.9)	.008
Febrile	14.3% (0.0-28.9)		28.1% (15.1-41.1)		26.1% (35.7-48.6)		41.2% (10.8-71.6)		20.8% (0.5-41.1)	.502
Tachycardic	64.3% (30.6-97.9)		70.3% (57.1-83.5%)		82.6% (63.1-100)		70.6% (42.4-98.8)		58.3% (33.7-82.9)	.490
Tachypnic	50.0% (14.9-85.1)		31.3% (17.8-44.7)		26.1% (3.6-48.6)		47.1% (16.2-77.9)		25.0% (3.3-46.7)	.336
Hypoxic	45.5% (4.0-86.9)		12.5% (2.9-22.1)		13.0% (0.0-30.3)		17.6% (0.0-41.2)		8.3% (0.0-22.2)	.054
SOS	13.7 (10.6-16.8)		14.1 (7.2-11.0)		14.9 (12.8-16.9)		14.9 (12.5-17.3)		13.0 (11.1-15.0)	.508
Duration	13.4 (7.0-19.8)		9.1 (7.2-11.1)		10.1 (7.1-13.0)		10.5 (4.8-16.3)		7.5 (5.5-9.4)	.159
Admitted	92.9% (74.8-100)		16.5% (6.0-26.9)		4.3% (0.0-14.8)		29.4 (1.2-57.6)		18.5% (0.4-36.6)	b.001

The 3 groups of seasonal influenza, H1N1 influenza, and other infection were combined into the viral all group. Means with 95% confidence intervals are depicted. Abbreviations: PTA, prior to arrival; SOS, Severity of symptoms score.

A IL-4

30

Cytokine Concentration pg/ml

B

150

Cytokine Concentration pg/ml

IL-5

20 100

10 50

0 0

GM-CSF

C D

Cytokine Concentration pg/ml

Cytokine Concentration pg/ml

INFy

100 1500

80

1000

60

40 500

20

0 0

E

Cytokine Concentration pg/ml

1000

IL-6

800

600

400

200

0

Fig. 1. Group cytokine concentrations. Patients are divided into 3 groups: (1) seasonal influenza, (2) H1N1 influenza, (3) bacterial pneumonia (PNA), and (4) other infection. Cytokines depicted are broken down into IL-4 (A), IL-5 (B), GM-CSF (C), IFN? (D), and IL-6 (E). Stacked bars are means with error bars depicting 95% confidence intervals. Statistical

significance highlighted by star (?).

A or B or the novel pandemic H1N1 virus. Viral culture results were used to group patients into the appropriate influenza infection groups. Patients not diagnosed with a bacterial pneumonia, whom had viral cultures negative for influenza virus, were placed into the other respiratory infection group. All 3 viral groups (seasonal, H1N1, and other) were also grouped together during analysis.

The primary outcomes of interest were diagnosis of bacterial infection and diagnosis of influenza infection, as described above. Hospital length of stay was determined by chart review.

Analysis

Subject demographics are presented as means with 95% confi- dence intervals. We used analysis of variance between groups to analyze differences in study group demographics. We compared the value of each cytokine concentration between bacterial pneumonia patients and viral-infected patients, using the Wilcoxon rank sum test. We calculated receiver operating characteristic curves for each cytokine and evaluated the predictive value in cases of bacterial infection. The cutoff points for each cytokine as a test for bacterial infection were determined by using the ROC curves, searching for the best sensitivity without significantly decreasing the specificity.

Results

Characteristics of study subjects

A total of 250 patients were screened over 3 influenza seasons, October through March in the years of 2008 through 2011. A total of

172 patients were enrolled. Of those patients, 54 (31.4%) were excluded from final analysis, leaving a final total of 118 patients. Subsequent exclusions were due to the following reasons: loss to follow-up (61.1%), use of immunosuppressant medications (3.7%), active liver disease (3.7%), inability to obtain a nasal wash sample (5.6%), inability to draw blood (1.8%), and diagnosis of another bacterial process (24.1%). Of the remaining 118 patients, a total of 80 (67.8%) were randomly chosen for analysis. This was accomplished through a random Selection process where the research assistant was blinded to all patient outcome information. Only 80 subjects were selected for analysis due to (1) limited availability of space to run samples within the assay system chosen and (2) the cost of the cytokine assay. We had sufficient funding to run only 160 samples and chose to run 80 subjects at 2 dilutions to obtain accurate results.

The selected 80 patients were divided into 4 groups based on the viral culture results and diagnoses of bacterial infection. There were

14 patients in the bacterial pneumonia group, 23 patients with seasonal influenza infection, 17 patients with novel H1N1 influenza virus infection, and finally 26 patients in the other respiratory infection group. This other respiratory infection group was defined by a normal CXR, negative viral culture, discharge diagnosis in the ED of viral infection, and absence of any complications or bacterial infection reported upon telephone survey follow-up and chart review. There were no patients diagnosed with bacterial pneumonia who had a positive viral culture for influenza. There were differences noted in these groups with regard to age, vaccine history, days of symptoms, and percentage of subjects admitted to the hospital (Table 1). The bacterial infection group was older compared with the other infection groups combined (P b .001); however, the average age of all subjects was relatively young, at 35.1 years of age. The bacterial group also had a longer duration of symptoms before presentation, at 4.1 days as compared with 2.4 days for viral infection groups (P = .008), and was more frequently admitted to the hospital, with only 1 patient being discharged from the ED (P b .001). Of note, all subjects scored their symptoms similarly with no observed differences in the severity of symptom scores. In addition, all groups had similar presenting characteristics of being febrile, tachycardic, tachypnic, and hypoxic.

There was a nonsignificant statistical trend in the bacterial group toward being hypoxic. All groups had a similar total duration of illness upon follow-up.

Main results

In the multiplex cytokine assay, the following cytokines showed differences between the groups analyzed: IL-4, IL-5, IL-6, IL-10, IFN?, and GM-CSF. The cytokines IL-1?, IL-2, IL-8, and TNF? showed no statistical significant differences between the 4 patient groups (data not displayed). Fig. 1 shows the averages for each of the 4 patient groups, with 95% confidence intervals. Patients with bacterial pneumonia had significantly lower sera concentrations of IL-4 (P =

.0005), IL-5 (P = .005), IFN? (P b .001), and GM-CSF (P = .001), while

demonstrating elevated concentrations of IL-6 (P = .006), when compared with both of the influenza infection groups and the other respiratory infection group. When considering all viral infections as a whole, there were significant differences in comparison with the bacterial pneumonia group with regard to IL-4 (2.66 vs 16.77 pg/mL, P b .001), IL-5 (20.57 vs 57.57 pg/mL, P = .006), IL-6 (403.06 vs 52.69 pg/mL, P b .001), GM-CSF (18.26 vs 66.80 pg/mL, P b .001), and IFN?

(0.0 vs 830.36 pg/mL, P b .001). There was no difference between the bacterial pneumonia group and all other viral infection groups combined with regard to IL-10. Interestingly, both the seasonal influenza and the H1N1 influenza groups demonstrated higher sera concentrations of IL-10 as compared with the bacterial and other infection groups combined (Fig. 2; P = .019).

Receiver operating characteristic curves for IL-4, IL-5, IL-6, IFN?, and GM-CSF were calculated as a predictor of bacterial pneumonia (Fig. 3). Interferon ? performed the best, with an area under the curve of 0.941. This result is a result of all 14 bacterial pneumonia patients demonstrated undetectable sera IFN? concentrations, whereas pa- tients in the other 3 groups demonstrated concentrations that were detectable by the assay system. In decreasing order, each cytokine and the respective ROC area under the curve were as follows: IL-4 (0.898), GM-CSF (0.850), IL-5 (0.822), and IL-6 (0.580). In comparison, the

ROC area under the curve for WBC was 0.741. According to the ROC curve analysis, the cutoff points with the optimal sensitivity and specificity were as follows: IL-4 at 9 pg/mL, IL-5 at 30 pg/mL, IL-6 at 150 pg/mL, GM-CSF at 30 pg/mL, and IFN? at 0 pg/mL. A cutoff of

10.5 x 10³/mm³ was used for WBC. Operating characteristics are

shown in Table 2.

As stated above, sera IL-6 was elevated in the bacterial pneumonia group, although with a wide confidence interval range. However,

IL-10

150

Cytokine Concentration pg/ml

100

50

0

Fig. 2. Interleukin 10 cytokine concentrations. Increased levels of IL-10 among both influenza groups (?) compared with bacterial pneumonia (PNA) and other respiratory infections. Stacked bars are means with error bars depicting 95% confidence intervals.

100

80

Sensitivity%

60

40

20

0

0 20 40 60 80 100

100% – Specificity%

WBC INFg GM-CSF IL6

IL5 IL4

2000

1500

> 3 Days Cytokine Concentration pg/ml

1000

500

0

IL6 Among Pneumonia Patients

15

< 3 Days Cytovkine Concentration pg/ml

10

5

0

Fig. 3. Receiver operator characteristic curves for total WBC and cytokines IFN?, GM- CSF, IL6, IL5, and IL4.

among bacterial pneumonia patients, those that were either dis- charged from the ED (n = 1) or had a hospital stay less than 72 hours had significantly lower sera IL-6 concentrations than those that had greater than 72-hour stays (Fig. 4). Patients admitted for more than 3 days, on average, demonstrated sera IL-6 concentrations of 702.67 pg/mL as compared with concentrations observed in admissions shorter than 3 days, of 3.59 pg/mL (P = .078).

Limitations

Our study has several limitations, including the selection process for the 80 patients enrolled and analyzed. We randomly selected 80 patients from the 118 enrolled subjects by blinding the sample identification numbers and choosing at random the samples to be further analyzed. This randomized selection was due to financial considerations and the cost of the cytokine panel assay system.

In addition, this study defines cases of bacterial pneumonia based on the presence of infiltrate seen on CXR coupled with discharge diagnosis. The diagnosis of bacterial pneumonia is a frequent clinical issue that has eluded the development of definitive evidence-based criteria despite many years of effort. The clinical definition criterion standard for bacterial infection diagnosis, which includes CXR infiltrate and positive Bacterial culture, is problematic because many patients with bacterial pneumonia have negative culture results, and many patients with infiltrate on chest radiograph do not have bacterial infection. Therefore, this study should be viewed as a promising preliminary investigation into the utility of sera cytokine concentrations in predicting cases of bacterial pneumonia. A much larger follow-up study is necessary to elucidate the utility and test characteristics of these cytokines as a diagnostic aid in cases of bacterial pneumonia in which the CXR is normal or equivocal. In addition, larger powered studies, including subgroup analysis of culture-proven bacterial pneumonia, in either sputum or blood, would aid in better defining the test characteristics of these cytokines. A further limitation of this study is that the bacterial infection group was significantly older, by nearly 20 years, than the other 3

Fig. 4. interleukin 6 levels among bacterial pneumonia patients who spent less than 72 hours (b3 days) vs greater than 72 hours (N 3 days) in the hospital. Stacked bars are means with error bars depicting 95% confidence intervals. The y-axis for the more than 3 days group is to the left, whereas the y-axis for the less than 3 days group is to the right.

groups, which may reflect 1 of 2 possibilities: (1) older adults are frailer and thus more susceptible to bacterial infection or (2) there is the possibility of intervention bias on the part of the treating ED clinician, who might be more apt to make a bacterial infection diagnosis in this age group regardless of CXR findings. Another possibility is that lower cytokine concentrations exist among adults who are older regardless of type of infection. We doubt this given that IL-6 concentrations are not affected by age [19], and we would not expect other cytokines to be preferentially affected over others by age. There is a nonsignificant trend of a greater percentage of women in the pneumonia group, which is likely due to the small sample size.

Finally, all but one of the pneumonia patients were admitted to the hospital. This leaves the appearance of a selection bias (ie, enrolling only pneumonia patients who have higher severity of illness and thus are being admitted). We attempted to limit this bias by enrolling patients before ED disposition and diagnosis. Larger studies are needed to address illness severity in bacterial pneumonia patients and the relationship of these cytokine concentrations.

Discussion

This investigation describes how a panel of cytokines may serve collectively as a marker of bacterial pneumonia in patients presenting with an ILI to the ED. The lower sera concentrations of IL-4, IL-5, GM- CSF, and IFN?, combined with an elevated sera IL-6, were associated with bacterial pneumonias and may serve as a diagnostic aid for patients presenting to the ED with symptoms of ILI. Conversely, patients presenting with ILI symptoms due to influenza or other viral etiologies were distinguished by low sera IL-6 and relatively elevated IL-4, IL-5, GM-CSF, and IFN?. Both seasonal and novel H1N1 pandemic influenzas were distinguished from unclassified bacterial or other infections by the elevation of the anti-inflammatory cytokine IL-10.

Table 2

Operating characteristics

	IL-4 (9 pg/mL)	IL-5 (30 pg/mL)	IL-6 (150 pg/mL)	GM-CSF (30 pg/mL)	IFN? (0 pg/mL)	WBC (10.5th/mm3)
Sensitivity	100% (73.2-100)	85.7% (56.2-97.5)	50.0% (24.0-76.0)	78.6% (48.8-94.3)	100% (73.2-100)	69.2% (38.9-89.6)
Specificity	76.5% (64.4-85.6)	67.6% (55.1-78.2)	97.1% (89.1-99.5)	80.9% (69.2-89.0)	88.2% (77.6-94.4)	71.1% (55.5-83.2)
PPV	46.7% (28.8-65.4)	35.3% (20.3-46.5)	77.8% (40.2-96.1)	45.8% (26.2-66.8)	63.6% (40.8-82.0)	40.9% (21.5-63.3)
NPV	100% (91.4-100)	95.8% (84.6-99.3)	22.2% (3.9-59.8)	94.8% (84.7-98.7)	100% (92.5-100)	88.9% (73.0-96.4)
LR	4.25 (2.77-6.52)	2.65 (1.77-3.97)	17.5 (4.05-75.6)	4.11 (2.35-7.20)	8.5 (4.43-16.30)	2.39 (1.34-4.30)

For the cytokines IL-4, IL-5, IL-6, GM-CSF, IFN?, and total WBC, the ideal cutoff points for each one as a test for bacterial infection was determined by using the ROC curves, searching for the best sensitivity without significantly decreased specificity. Each row depicts the sensitivity, specificity, positive predictive value, negative predictive value, and likelihood ratio for the selected cytokine and WBC using the depicted cutoff values. The 95% confidence intervals are in parentheses. Abbreviations: PPV, positive predictive value; NPV, negative predictive value; LR, likelihood ratio.

Each of the assayed cytokines, with the exception of IL-6, out- performed the clinically standard use of elevated WBC, in the diagnosis of bacterial pneumonia.

In the state of acute inflammation, local immune cells generate Proinflammatory cytokines including TNF?, IL-1?, IL-6, IL-8, GM-CSF, and IFN? [20]. In response to inflammatory cytokines, the liver produces acute phase proteins including CRP [5]. Anti-inflammatory cytokines such as IL-1 receptor antagonist, IL-4, IL-10, IL-11, IL-13, and transforming growth factor ? [21] act in concert with proinflamma- tory cytokines to mediate Acute infection and sepsis, a potential outcome of CAP [20]. In the state of natural infection, this milieu is difficult to interpret, as some cytokines, such as IL-6, have roles as both proinflammatory and anti-inflammatory effectors.

Based on our results, we can not only confirm the sera elevation of IL-6 as a diagnostic indictor of bacterial infection but also distinguish bacterial from viral infections based on IL-4, IL-5, GM-CSF, and IFN?. In addition to the valuable clinical application of these profiles, we may also infer information about the natural course of bacterial vs viral infection. In the case of patients presenting with symptoms of ILI, it appears that viral infection produces a more diverse immune response as compared with bacterial infection. Influenza or other viral infections in this study were characterized by both proinflam- matory and anti-inflammatory cytokines as well as cytokines associated with the generation of a humoral response, IL-4 and IL-5. The reason for this distinction is not known. It may be due to the mechanism of infection, as bacterial agents are extracellular patho- gens, whereas viruses operate intracellularly. Typically, bacterial invasion activates the complement cascade, whereas viral infections, such as influenza, do not.

Of the numerous cytokines that have been measured in patient sera during acute inflammation, IL-6 has been the best studied [19]. Elevated sera IL-6 concentrations have been shown in bacterial pneumonia to correlate with higher CURB-65 scores and predict poorer outcomes in CAP [22,23]. Our investigation not only confirms this cytokine as a marker of bacterial pneumonia but also demon- strates this elevation in comparison with seasonal and pandemic influenza infections and other infections. In this study, there was an observed correlation of elevated sera IL-6 concentrations with longer hospital stays. Throughout the literature, elevated IL-6 concentrations have been associated with mortality from CAP, whereas lower IL-6 concentrations were observed following successful antibiotic treat- ment [24,25]. Although IL-6 has been shown to be a marker of bacterial pneumonia, its relationship to duration of hospital stay has not previously been shown. Other cytokines, such as IL-8, have also been shown to be elevated in bacterial pneumonia [24]; however, this response has varied depending on the type of bacteria causing the pneumonia [26]. In this study, the type of bacterial pneumonia was not identified.

In this report, not only did our panel of cytokines differentiate bacterial from viral infections in cases of ILI, but these markers also served to distinguish influenza infection from both bacterial and other infections, based on the elevation of sera IL-10. To our knowledge, this has not been previously reported. We did not observe differences in the studied cytokines between seasonal and pandemic H1N1 influenza, which have been previously described [15]. This may be related to the relatively young patient population enrolled in this study.

In summary, the sera cytokines IL-4, IL-5, IL-6, IL-10, IFN?, and GM- CSF showed significant differences between bacterial, influenza, and other respiratory tract infections and may thus serve as potential guides for both the diagnosis and management of patients presenting to the ED with an ILI during the influenza season. Measurement of these cytokines may serve as a tool for the ED physician in prescribing specific antiviral vs antibacterial therapy or a combination of both. immunological markers to aid the ED physician in the diagnosis of

bacterial and influenza infections and in identifying those patients that may benefit from antibiotics, Antiviral medications, or hospital- ization would lead to a shift in clinical practice, reducing unnecessary antibiotic use and improving patient outcomes.

References

1. Basi SK, Marrie TJ, Huang JQ, Majumdar SR. Patients admitted to hospital with suspected pneumonia and normal chest radiographs: epidemiology, microbiolo- gy, and outcomes. Am J Med 2004;117(5):305-11.
2. Albaum MN, Hill LC, Murphy M, et al. interobserver reliability of the chest radiograph in community-acquired pneumonia. PORT Investigators. Chest 1996;110(2):343-50.
3. Lee JY, Hwang SJ, Shim JW, et al. Clinical significance of serum procalcitonin in patients with community-acquired lobar pneumonia. Korean J Lab Med 2010;30(4):406-13.
4. Ebell MH. diagnosing and treating patients with suspected influenza. Am Fam Physician 2005;72(9):1789-92.
5. Whicher JT, Chambers RE, Higginson J, Nashef L, Higgins PG. Acute phase response of serum amyloid A protein and C reactive protein to the common cold and influenza. J Clin Pathol 1985;38(3):312-6.
6. Falsey AR, Walsh EE, Francis CW, Looney RJ, Kolassa JE, et al. Response of C-reactive protein and serum amyloid A to influenza A infection in older adults. J Infect Dis 2001;183(7):995-9.
7. Haran JP, Beaudoin FL, Suner S, Lu S. C-reactive protein as a predictor of bacterial infection among patients presenting with an influenza-like illness during the influenza season. Am J Emerg Med 2013;31(1):137-44.
8. Summah H, Qu JM. Biomarkers: a definitive plus in pneumonia. Mediators Inflamm 2009;2009(675753):1-9.
9. Lehtomaki K, Leinonen M, Takala A, Hovi T, Herva E, Koskela M. Etiological diagnosis of pneumonia in military conscripts by combined use of bacterial culture and serological methods. Eur J Clin Microbiol Infect Dis 1988;7(3):348-54.
10. Ichiyama T, Morishima T, Isumi H, Matsufuji H, Matsubara T, et al. Analysis of cytokine levels and NF-kappaB activation in peripheral blood mononuclear cells in influenza virus-associated encephalopathy. Cytokine 2004;27(1):31-7.
11. Hayden F, Fritz R, Lobo M, Alvord W, Strober W, Straus SE. Local and systemic cytokine responses during experimental human influenza A virus infection. Relation to symptom formation and host defense. J Clin Invest 1998;101(3): 643-9.
12. Peiris JS, Poon LL, Guan Y. Emergence of a novel swine-origin influenza A virus (S-OIV) H1N1 virus in humans. J Clin Virol 2009;45(3):169-73.
13. Bautista E, Chotpitayasunondh T, Gao Z, Harper SA, Shaw M, Uyeki TM, et al. Clinical aspects of pandemic 2009 Influenza A virus infection. N Engl J Med 2010;362(18):1708-19.
14. Kobasa D, Takada A, Shinya K, Hatta M, Halfmann P, et al. Enhanced virulence of influenza A viruses with the haemagglutinin of the 1918 pandemic virus. Nature 2004;431(7009):703-7.
15. Monsalvo AC, Batalle JP, Lopez MF, Krause JC, Klemenc J, et al. Severe pandemic 2009 H1N1 influenza disease due to pathogenic Immune complexes. Nat Med 2011;17(2):195-9.
16. Treanor JJ, Hayden FG, Vrooman PS, Barbarash R, Bettis R, et al. Efficacy and safety of the oral neuraminidase inhibitor oseltamivir in treating acute influenza: a randomized controlled trial. JAMA 2000;283(8):1016-24.
17. Berthoud TK, Manaca MN, Quelhas D, Aguilar R, Guinovart C, et al. Comparison of commercial kits to measure cytokine responses to Plasmodium falciparum by multiplex microsphere suspension array technology. Malar J 2011;10:115.
18. Elshal MF, McCoy JP. Multiplex bead array assays: performance evaluation and comparison of sensitivity to ELISA. Methods 2006;38(4):317-23.
19. Groeneveld AB, Tacx AN, Bossink AW, van Mierlo GJ, Hack CE. Circulating Inflammatory mediators predict shock and mortality in febrile patients with microbial infection. Clin Immunol 2003;106:106-15.
20. Pinsky MR. Sepsis: a pro- and anti-inflammatory disequilibrium syndrome. Contrib Nephrol 2001;132:354-66.
21. Opal SM, DePalo VA. Anti-inflammatory cytokines. Chest 2000;117:1162-72.
22. Zobel K, Martus P, Pletz MW, Ewig S, Prediger M, Welte T, et al. Interleukin 6, lipopolysaccharide-binding protein and interleukin 10 in the prediction of risk and etiologic patterns in patients with community-acquired pneumonia: results from the German competence network CAPNETZ. BMC Pulm Med 2012;12(1):6, http://dx.doi.org/10.1186/1471-2466-12-6. BioMed Central Ltd.
23. Padrones S, Garcia-Vidal C, Fernandez-Serrano S, Fernandez a, Masuet C, Carratala J, et al. Impact of antibiotic therapy on systemic cytokine expression in pneumococcal pneumonia. Eur J Clin Microbiol Infect Dis 2010;29(10):1243-51, http://dx.doi.org/10.1007/s10096-010-0993-0.
24. Endeman H, Meijvis SC a, Rijkers GT, van Velzen-Blad H, van Moorsel CHM, Grutters JC, et al. Systemic cytokine response in patients with community- acquired pneumonia. Eur Respir J 2011;37(6):1431-8, http://dx.doi.org/10.1183/ 09031936.00074410.
25. Kellum JA, Kong L, Fink MP, Weissfeld LA, Yealy DM, Pinsky MR, et al. Understanding the inflammatory cytokine response in pneumonia and sepsis: results of the Genetic and Inflammatory Markers of Sepsis (GenIMS) study. Arch Intern Med 2007;167(15):1655-63.
26. Kragsbjerg P, Jones I, Vikerfors T, Holmberg H. Diagnostic value of blood cytokine concentrations in acute pneumonia. Thorax 1995;50(12):1253-7.