Background
Community-acquired pneumonia (CAP) is the most common potentially fatal infectious disease in industrialized countries [
1,
2]. In order to classify pneumonia severity and to guide treatment decisions, several prognostic scores for CAP have been developed. The CURB-65 and CRB-65 scores are currently the preferred scores in Europe because of their simplicity and applicability. The CRB score has been shown to be less useful for the estimation of the risk of death within 30 days [
3,
4].
A variety of pro- and anti-inflammatory cytokines are involved in the regulation of the inflammatory response to pulmonary infections; measurements of these cytokines have been used to improve diagnosis and assess severity, and thus predict the prognosis of CAP. The concentrations of interleukin 6 (IL-6) [
5] and interleukin 10 (IL-10) [
6] reflect inflammatory and anti-inflammatory reactions, respectively. The concentration of lipopolysaccharide-binding protein (LBP) is increased during bacterial infections [
7].
IL-6 is secreted by a variety of cells such as T-lymphocytes, macrophages, endothelial cells and epithelial cells. In inflammatory reactions the rise of IL-6 concentration precedes that of most of the other cytokines [
5,
7‐
9] and it has a plasma half-life of 20-60 min. IL-10 is an anti-inflammatory cytokine that is produced primarily by monocytes and to a lesser extent by Th-2 lymphocytes. IL-10 inhibits the synthesis of pro-inflammatory cytokines and suppresses antigen presentation, and normal serum half-life is only a few hours [
6]. LBP is an acute-phase protein that is increased 30-fold within 24-48 h of infection [
7] and it forms complexes with bacterial lipopolysaccharides. These complexes bind to specific receptors on monocytes and macrophages and boost inflammation. The serum half-life of LBP is 2-4 h.
The aim of our study was to investigate the predictive value of IL-6, IL-10 and LBP compared with the CRB-65 score with respect to the clinical course of CAP and the microbial etiology.
Methods
Study population
CAPNETZ is a German competence network for the study of CAP. It collects data and samples (blood, urine and sputum) from patients with proven CAP (hospitalized patients and outpatients). The network consists of local clinical centers throughout Germany [
1].
Samples and patient data were obtained from CAPNETZ study group. All cases of CAP were reported, via a network of sentinel practices and hospitals, to the study monitor of the corresponding local clinical center.
The recruitment period for the present study was January 2001 to December 2006. A total of 3,220 subjects were selected from the database. The population included 105 subjects with a severe course of CAP, which was defined as mechanical ventilation, intensive care unit (ICU) treatment and/or death within 30 days of involvement. These 105 subjects, and a random sample of 895 controls stratified for CRB-65 score, were included in the study. The control group was defined using the algorithm described as follows. First, the total number of evaluable cases was determined, n = 105. The total size of the study population was chosen in order to get 1,000 probes for analysis. Thus with 105 cases, 895 controls had to be selected. For the selection of the control sample population, the number of subjects with CRB-65 0-3 was fixed to match the frequency distribution of CRB-65 in all controls. Then, the respective controls were selected from all controls in each CRB-65 stratum using the random number generator of SPSS software (SPSS, Chicago, Illinois, USA).
Samples (blood, sputum and urine) were taken within 24 h of first presentation and inclusion in the CAPNETZ study.
The inclusion criteria were: age ≥ 18 years, a new pulmonary infiltrate diagnosed by chest radiograph, written informed consent prior to inclusion in the study, and at least one of the following clinical symptoms: cough with purulent sputum, fever or positive auscultation.
The exclusion criteria were: acquired or therapeutically induced immune deficiency, active tuberculosis, or hospitalization less than 4 weeks prior to infection.
The decision where (inpatient or outpatient) and how to treat the patient was made by the attending physician.
All patients were assessed at first presentation and during follow-up according to a standardized web-based data sheet. After 14, 28 and 180 days, all patients were contacted either in person or via telephone for a structured interview on outcome parameters (for example: resolution of symptoms, length of antimicrobial therapy or death) [
10,
11].
The CAPNETZ project was approved by the local ethics committee. Written informed consent was obtained from each patient prior to his/her inclusion in the study.
Determination of CRB and CRB-65 scores
For all patients the CRB and CRB-65 scores were determined on inclusion in the study [
10]. The CRB score consists of three variables: confusion, respiratory rate ≥ 30/min, and systolic blood pressure < 90 mmHg or diastolic blood pressure ≤60 mmHg. The CRB-65 score adds age ≥ 65 years to the CRB score, and one point is given for each parameter present. CRB scores range from 0 to 3, and CRB-65 scores range from 0 to 4.
Measurement of IL-6, LBP and IL-10
All serum samples to be tested for IL-6, LBP and IL-10 were stored at -70°C. The cytokine concentrations were measured using a solid-phase, two-site chemiluminescence immunometric assay on Immulite 1000 (LBP and IL-10) or Immulite 2000 (IL-6; Siemens Healthcare, Eschborn, Germany) according to the manufacturer's instructions. The detection limits are 1.2 μg/ml, 0.99 pg/ml and 2 pg/ml for LBP, IL-10 and IL-6, respectively.
Microbiologic diagnosis
Potential pathogens were detected using microbiologic, serologic and molecular biological standard procedures, as described previously [
11,
12].
Streptococcus pneumoniae, Haemophilus influenzae, Klebsiella pneumoniae, Escherichia coli and other enterobacterial species,
Pseudomonas aeruginosa, Moraxella catarrhalis and
Stenotrophomonas maltophilia were regarded as definite causative pathogens when they were isolated from: (1) blood cultures or pleural fluid cultures, or (2) good-quality sputum containing < 25 polymorphonuclear cells and < 10 epithelial cells per high-power field (total magnification × 1,000), and there was predominant growth from culture of the sputum of > 10
6 colony-forming units (CFU)/ml or > 10
4 CFU/ml from bronchoalveolar lavage fluid. If one of these criteria was not met, the presence of any of the bacteria listed above was judged to be indeterminate. In addition,
Chlamydia pneumoniae (IgM positive ≥ 1:32 and/or positive PCR),
Legionella spp. (culture, positive urinary antigen and/or PCR),
Mycoplasma pneumoniae (PCR),
S. pneumoniae (urinary antigen), influenza A and B viruses, respiratory syncytial virus, adenovirus, enterovirus (all PCR) and
Aspergillus spp. (only with concomitant lung abscess and/or histologic confirmation) were considered to be causative pathogens.
Statistics
Statistical analyses were performed with SPSS software. Two-group comparisons of non-parametric data were calculated by using a t-test. For multigroup comparisons, one-way ANOVA with Bonferroni post-hoc analyses were used. No correction for multiple testing was applied with respect to the three cytokines (IL-6, IL-10 and LBP). The operating characteristics of IL-6, IL-10 and LBP were assessed by calculating sensitivity, specificity and predictive values.
We constructed receiver-operating characteristic (ROC) curves and determined the area under the curve (AUC). The combination of parameters was done using the linear predictor obtained from logistic regression analysis. Thus, no categorization was applied. The usefulness of the markers for improving the prediction by CRB-65 score was evaluated by calculation of continuous net reclassification improvement (NRI) as described by Pencina
et al. [
13,
14]. Thus, we looked at the rates of increased and decreased event probability for each subject without applying a categorization.
P values below 0.05 (two-sided) were considered to be statistically significant. A comparison of ROC curves was done by using the method of DeLong [
15], which takes into account the correlation of measurements and increases the power considerably, compared with the 'naïve' method, which uses raw standard errors.
Samples from patients who did not have a severe course of CAP were used for all controls, but the percentage of patients with a severe course of CAP was about threefold (10.5% versus 3.3%) more in the final study population as compared with that in the primary sample of 3,220 subjects. Thus, sensitivity and specificity of scores could be determined without further correction; however, predictive values had to be corrected using the formula of Bayes.
Discussion
This study demonstrated: (1) IL-6 and IL-10 concentrations at the time of patient admission reflected, similar to the CRB score, the severity of disease rather than prognosis, which is influenced not only by severity but also by age and co-morbidities of the patients; therefore, high IL-6 and/or IL-10 concentrations were found in patients with high CRB or CRB-65 scores; (2) the biomarkers used were inferior to CRB-65 score, but similar to the CRB score, for predicting a severe course of CAP; (3) the accuracy of both CRB and CRB-65 for predicting a severe course of pneumonia could be increased significantly by including the IL-6 measurements in a statistical model; (4) LBP had some predictive potential in etiology, but individual predictions could not be made.
Current CAP guidelines recommend that patients at increased risk of death should be hospitalized, and that ICU admission should be considered for patients with the highest risk. To date, several different pneumonia severity indexes, such as the pneumonia severity index, CURB, CURB-65, CRB and CRB-65, have been evaluated [
16‐
18]. In addition, new biomarkers such as procalcitonin (PCT), proatrial natriuretic peptide (pro-ANP, copeptin), proatrial vasopressin (proAVP), proadrenomedullin (proADM) and proendothelin-1 (proET1) have been found to have similar, additive or superior predictive values for the risk assessment of CAP [
19,
20].
IL-6 and IL-10 play an important role in regulation of inflammatory reactions. Various studies have shown increased concentrations of these cytokines during pneumonia [
21‐
24]. Elevated IL-6 and IL-10 levels have been directly associated with higher mortality, ICU admission and mechanical ventilation [
25,
26].
Increased LBP concentrations have been found in patients with severe sepsis caused by Gram-negative or Gram-positive bacterial infections or fungal infections [
25,
26]. Data concerning LBP concentration in CAP are scarce [
27,
28].
In contrast to most of the studies that have been published so far, we investigated the concentrations of IL-6, IL-10 and LBP in comparison with clinical and microbiologic findings in a large and carefully characterized study population, which included a large number of patients with a severe course of CAP; this was defined as mechanical ventilation, ICU treatment and/or death within 30 days of involvement.
In patients with a severe course of pneumonia, we found significantly increased IL-6 and LBP concentrations. In contrast to previous studies, which found new biomarkers such as PCT, pro-ANP, proAVP, proET1 and proADM to be good predictors of the course and the outcome of CAP, and which were similar or superior to CRB-65 score, we found a lower predictive potential for IL-6, IL-10 and LBP. As published recently, the predictive value of CRP was also lower than that of the CRB-65 score [
10]. We were able to confirm these findings by calculating an AUC of 0.663 for differentiation of the combined end point and CRP (data not shown). The combination of CRB-65 score and IL-6 concentration increased the predictive value. This finding may be of limited value for the risk stratification of individual patients as there was a significant overlap in the distribution of IL-6 concentration in different patient groups [
29]. Therefore, to test the value of IL-6 for the improving risk prediction we calculated the NRI for IL-6 in combination with CRB-65 score and compared it with CRB-65 score alone and found a significant net gain in reclassification by including IL-6.
On the other hand, we found a strong association between the IL-6 and IL-10 concentrations and high CRB scores (class 2-3). The operational characteristics of IL-6, IL-10 and CRB with respect to a severe course of CAP were similar to each other.
There are several factors that may reduce the prognostic potential of these parameters. First, the induction of IL-6 secretion precedes that of most other pro-inflammatory cytokines and is strongly controlled by anti-inflammatory cytokines such as IL-10. Second, all cytokines investigated in our study have a short plasma half-life of a few hours. Previous studies have shown that the mean levels of IL-6 and IL-10 decrease rapidly after clinical presentation in response to therapeutic interventions [
30‐
32]. Thus, IL-6 and IL-10 are cytokines that characterize the acute clinical condition of patients. Therefore, and with respect to the high negative predictive values of 0.984, 0.981 and 0.978 for IL-6, IL-10 and LBP, respectively, normal cytokine concentrations might be used to identify patients who can be treated as outpatients.
Similar to our results, previous studies have shown that the CRB score fails to predict many severe courses of CAP. Consequently, it has been suggested that age ≥ 65 years should be included to identify patients at high risk because of potential comorbidities [
25,
26]; the resulting CRB-65 score has been shown to useful in characterizing the acute clinical condition and the risk of a severe course of CAP.
Cardiovascular markers such as copeptin (proAVP), proET1 or MR-proANP, or adrenomedullin (MR-proADM) which has multiple effects, reflect at least in part non-pulmonary comorbidities [
19,
20]. These biomarkers have the highest predictive value for death within 30 or 180 days. Inflammatory biomarkers such as PCT, which has a delayed concentration kinetic, have an intermediate predictive value that is comparable with the clinical CRB-65 score. We have shown that inflammatory markers that are rapidly regulated have a low predictive value that is comparable with that of the CRB score. Further studies should answer the questions: (1) does a combination these biomarkers with a cardiovascular marker further improve the prediction of the course of CAP, and (2) can these biomarkers be used to monitor disease progression? The latter question is not addressed within the CAPNETZ study because the timing of the biomarker measurements is not standardized, except that the blood samples were collected within 24 h after presentation at the study center. Under these conditions biomarkers that have short peak concentrations at defined disease states may have lower prognostic values compared with those with long-lasting elevated concentrations.
In CAP, the determination of the etiology of the microbial pathogen is essential to target therapeutic decisions. In our study population the etiology remained unknown in 67.9% of the patients; this is comparable with previous findings [
11]. It has been suggested that diagnostic sensitivity is reduced by antibiotic pretreatment prior to hospital admission. For example, typical pathogens such as
S. pneumoniae are easily missed after a single dose of antimicrobial treatment [
11,
33]. In our study we found no association between antibiotic pretreatment and identification of etiologic microbial pathogen. The identification rate was higher patients with a severe course of pneumonia.
The highest concentrations of LBP, IL-6 and IL-10 were found in patients infected with typical bacteria. In previous studies, elevated IL-6 concentrations were found in patients infected with
S. pneumoniae [
24,
34]. The concentration of the anti-inflammatory IL-10 was increased in parallel to the increase in pro-inflammatory IL-6, and this reflects the amplification of antimicrobial as well as host-protective processes in CAP [
35,
36].
In our study, the LBP concentration was the best parameter to differentiate between typical and atypical or viral pathogens, and we found significantly higher LBP concentrations in patients with typical bacterial infections. In addition, in our study population, we found significantly higher LBP concentrations in patients with atypical infections compared with patients infected with an unknown pathogen. Masia
et al. found LBP concentrations to be similar in patients infected with unknown pathogens, compared with those with viral and atypical infections [
37]. Within the group of typical bacterial infections, we found higher LBP concentrations in patients infected with Gram-positive bacteria. This supports the previous finding that LBP is a non-specific marker of bacterial infection [
25,
26].
Conclusions
In conclusion, we have shown that IL-6, LBP and IL-10 are biomarkers that are significantly increased in severe CAP. Increased IL-6 and IL-10 concentrations reflect an acute inflammation at time of admission and are associated with high CRB and CRB-65 scores. The predictive value for a severe course of CAP, defined by death, admission to ICU or mechanical ventilation, was lower than that of the CRB-65 score. However, the accuracy of both CRB and CRB-65 scores could be increased significantly by including the IL-6 measurements in the statistical model. IL-6 and IL-10 represent, with respect to CAP, a group of biomarkers that reflect the current clinical condition (that is, severity of disease) of patients; this is different from previously studied biomarkers, such as pro-ANP or proADM, which reflect cardiovascular comorbidities, and PCT, which might reflect persisting inflammatory reactions. LBP appeared to be the best parameter to predict a typical bacterial infection. Further studies should address the question of whether the predictive value can be improved by combined measurement of biomarkers from different groups, and whether these cytokines are useful biomarkers to monitor disease progression and response to therapeutic interventions.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
KZ and FB carried out the immunoassays, part of the statistical analyses and drafted the manuscript. PM performed part of the statistical analysis. TW and MP participated in its design and coordination. SE and MP participated in the design of the study and helped to draft the manuscript. All authors read and approved the final manuscript.