Our study showed that the detectable levels LBP, IL-6 and CRP differ in the earl stage of infection dependent on severity of infection. Furthermore, the detectable levels of these biomarkers differ in children with bacteraemia and those without. In addition, our results showed that LBP, IL-6 and CRP can be used as biomarkers in early detection of bacteraemia.
To date only a few studies are available on the association between SIRS and sepsis in children as the definitions of SIRS and sepsis in children have only recently been adopted.
The inflammatory biomarkers - CRP, IL-6 in children population have been studied comparatively widely, clinical studies of LBP in children are limited and to our knowledge HMGB1 have not been studied in infected children population [14–18]. The previous studies in children population are performed in very restricted patients' populations - mainly neonatal age, which have very age specific physiological conditions with possible influence on results; severe ill patients from intensive care units. Our study, however, included children from all-age groups covering a large spectrum of children with different grades of infections.
The second aim of our study was to facilitate the diagnosis of bacteraemia, as the recognition process of pathogen in the blood is still not perfect. Frequently bacteraemis is strongly suspected without microbiological confirmation and physician institutes the treatment as for bacteraemeic patient. To improve the diagnosis of possible bacteraemia by inflammatory markers, and speed the sequential start of appropriate treatment, we include patients with the high suspicions of bacterial infection (probable bacteraemia) in the group of bacteraemic patients. It is possible that besides gold standard - positive blood culture, the inflammatory biomarkers could speed the diagnosis of bacteraemia. The consequential of it, is the heterogeneity of the bacteraemic patients group, which could be evaluate as the drawback.
HMGB1 has been known from many years as chromosomal protein, but in recent years it has been very intensively investigated as a proinflammatory cytokine [9, 23]. In 1999, Wang et al  found increased levels of HMGB1 in 25 critically ill patients with sepsis, and significantly higher levels in those that succumbed to the disease. Similar findings were reported by Hatada et al.  who detected moderately elevated HMGB1 levels in patients with infectious diseases and highest HMGB1 levels in patients who died. Gaini et al.  maintained that HMGB1 levels failed to discriminate between internal medicine department patients with infection and those without infection, but HMGB1 levels were significantly higher in patients compared with healthy controls. Similar findings were reported by Sunden-Cullberg et al. , who noted high HMGB1 levels in patients with sepsis and septic shock, but found no correlation between HMGB1 concentration and the severity of illness. Angus et al.  found that HMGB1 levels did not differ between those with and without sepsis. One possible explanation for these divergent results could be the different laboratory methods used (Western blot and ELISA). Several authors [8, 27] have stated that in studies where blotting methods were used, higher levels of HMGB1 in range from 84 ng/ml to over 340 ng/ml were observed [9, 25, 26]. Using HMGB1 ELISA techniques, markedly lower HMGB1 levels were found. Hatada et al.  observed median HMGB1 levels of 4.5 ng/ml] in infected patients. HMGB1 levels of 7.7 ng/ml were measured by Yasuda et al.  in infected patients with severe pancreatitis. Gaini et al,  in infected patients without sepsis in one of their studies noted median HMGB1 levels of 2.41 ng/ml, and median levels of 3.4 ng/ml in another study  which is comparatively close to our HMGB1 result in this patients' group - 2.5 ng/ml. Mean levels of HMGB1 in sepsis patients in our study was 3.0 ng/ml, which are in line with the HMGB1 levels in sepsis patients found in the studies by Gaini et al [10, 12]. Furthermore, we did not found a significant difference between HMGB1 levels in infected patients without SIRS and patients with sepsis. These findings are in line with those from Gaini et al  as well. In our study using an ELISA technique, median HMGB1 levels were 2.5-3.2 ng/ml which correlate with the lower HMGB1 levels observed in others studies where ELISA technique was used. In contrast to another study , we did not found a significant difference between HMBG1 levels in bacteraemic and non bacteraemic patients. This may be due to the fact that in our study not only bacteraemic patients, but also patients with possible bacteraemia were considered as having bacteraemia introducing a misclassification bias. Another possible explanation may be the different age of the children in our study compared to the one conducted by Gaini et al. The HMGB1 is a "late onset" proinflammatory cytokine, but our population was sampled at the early stage of disease. This may be a reason for not finding a significant difference in HMBG1 levels in our study groups.
LBP is a comparatively recently identified protein synthesized mainly in liver. It binds to lipopolysaccharide of Gram negative bacteria, and elevated levels have been seen in infections caused by Gram positive bacteria both in adult [29, 30] and children populations . We found a moderate correlation between LBP and CRP, which concurs with the findings of Gaini et al.  and partly with their findings in their other publication where a strong correlation was found . We saw a moderate correlation between LBP and IL6, which was also found by Gaini and collegues. In their other study , these authors noted a weak correlation between LBP and IL6. In our study LBP levels in sepsis patients (median 26.4 μg/ml) correlates very closely with data from Pavcnik-Arnol  where LBP concentrations in SIRS patients with sepsis from population of critically ill neonates were 27.1 μg/ml. Gaini et al. in their studies in sepsis patients accordingly observed LBP levels of 33.5 μg/ml and 63.3 μg/ml [10, 12]. In our study patients with severe sepsis median LBP level was 79.7 μg/ml which correlate with Gaini et al.  results - LBP levels in severe sepsis patients were 88.7 μg/ml. The levels of LBP were significantly different among infected patients with SIRS and sepsis patients, as well among severe sepsis and sepsis patients which correspondent to the findings from Gaini et al. studies [10, 12, 31]. In our study LBP levels were significantly different among patients with bacteraemia and without bacteraemia, these results are in line with findings by Gaini et al. . In our study LBP with an AUC of 0.82 performed well on ROC analysis to examine diagnostic abilities in detecting bacteraemia, which correlates with the study by Pavcnik-Arnol et al.  where AUC values for LBP in diagnosis of bacterial sepsis were 0.82 in older children, 0.93 in neonates over 48 h and 0.97 in neonates aged under 48 h. LBP in Gaini at al.  study in adult patients population with AUC of 0.74 did not perform well in ROC analysis examining its ability to identify bacteraemic patients. In our study using a cut-off level of 26.6 μg/ml, LBP had a sensitivity of 80% and a specificity of 55.4% in diagnosis of bacteraemia which rather correlates with a cut-off level for LBP of 20 μg/ml with 91% sensitivity and 85% specificity by Pavcnik-Arnol  in neonates <48 h of age and cutoff 13.3 mg/l for children age.
The levels of the inflammatory cytokine IL6 in our study showed statistically significant differences between infected patients without SIRS and sepsis patients: IL6 was markedly higher in sepsis patients than in patients without SIRS. Overall these results match those of other studies analyzing the levels of IL6 in infected patients without SIRS and sepsis in surgery patients, neonatal age patients, children with acute apendicitis [33–36]. The median levels of IL6 were significantly different among infected patients with SIRS and sepsis patients, and among severe sepsis and sepsis patients, these results correlate Gaini et al. studies [10, 12, 31]. Median IL6 level for severe sepsis patients in our study was 290 pg/ml, which correlate with the median IL6 levels from Gaini et al. , where medium IL6 levels were 199.3 pg/ml for severe sepsis. We found significant difference in IL6 levels among bacteraemic and non bacteraemic patients. IL6 levels 21.2 pg/ml and 178.1 pg/ml accordingly were detected, Gaini et al.  observed accordingly 50.3 pg/ml and 178.1 pg/ml in the same patients' groups. We detected a moderate correlation between IL 6 and LBP which is partly consistent with findings from other studies, where in one study a weak correlation was detected and in another a moderate connection was fixed [12, 31]. In diagnosing bacteraemia we obtained a cut-off level of 58.7 pg/ml for IL-6 with a sensitivity of 80% and a specificity of 55.4%. Our result is in contrast to results by Pavcnik-Arnol  who reported a cut-off of 43.2 pg/ml for bacterial neonatal sepsis diagnosis and Silveira et al.  who found an optimal cut-off of 32 pg/ml for neonatal sepsis diagnosis on one side, and a cut-off value of 94.6 pg/ml from Gaini et al.  in diagnosing bacteraemia in adult population from another side. In our study on ROC analysis IL-6 performed good, with AUC 0.87, which is quite close to results of Groselj-Grenc  - AUC 0.776 for acute appendicitis diagnosis in children, and marginally different from results in Pavcnik-Arnol  study - AUC 0.67 for prediction of bacterial sepsis in neonates.
The statistically significant differences in median CRP levels observed between the three groups of children with different severity of infection were in line with the findings of other studies showing higher CRP levels with higher severity of infection in children [31, 37].
Naturally, our study had some limitations. The number of patients was rather small due to refusal of study participants (29 patients refused to participate) and the lack of a control group. Furthermore, we included patients with strong suspected (but not confirmed) bacteraemia into the group with bacteraemia which may have caused misclassification bias. However, in clinical work this group would be considered as having bacteraemia and would receive adequate treatment reflecting the real situation in clinical practice.
The issues to be improved are the number of patients included in the study, necessity of a control group consisting of healthy children, and the inclusion of children in the newborn age group. We have recognized the potential problems which we have to meet in future: difficulties to establish the control group - to receive the agreement from parents to take intravenous blood samples from healthy children will be complicated; difficulties to include newborn children in study due to local health-care system (newborn treatment is mainly organized in specialized centers at maternity departments). We have focused on research results that may be useful in future clinical practice for early severe infections diagnosis in children by using effective diagnostic markers.