Sepsis is still the main cause of death in surgical intensive care units with a continuously increasing incidence and a mortality rate ranging from 30% to 60% depending on sepsis severity and the days of hospital stay ^[1], ^[2]. Therefore, both an early diagnosis and prognosis of sepsis are of utmost importance to control efficacy of antibiotic and surgical therapy, to manage further diagnostics and interventions and to optimize cost containment. To identify patients that are at high risk to succumb to sepsis, however, is difficult, due to a vast variety of influencing factors (e.g. age, underlying disease, co-morbidity, focus and type of infection, readiness and adequacy of therapy). Clinical and routine labarotory signs, like fever and leukocytosis, respectively, as well as clinical scores like APACHE II and SAPS II are not always helpful for outcome prediction ^[3], ^[4]. Therefore biomarkers that are released during the inflammatory response, like PCT, IL6 and CRP, have been investigated and suggested as useful parameters to determine the outcome of septic patients ^[4], ^[5], ^[6], ^[7], ^[8], ^[9], ^[10], ^[11].

However, there is no generally accepted marker for monitoring the evolution of sepsis. More recently, Lipopolysaccharide-binding protein (LBP) has been proposed as a sensitive marker for bacterial infection ^[12], ^[13] and possibly useful follow-up parameter in detection and resolution of sepsis ^[14], ^[15]. Like CRP, LBP is an acute phase protein, that is produced by hepatocytes as well as epithelial cells of the intestine and the lungs ^[16], ^[17] after induction by interleukin-(IL)-6 and IL-1, and a key participant in the innate immune response to Gram-negative, Gram-positive bacterial and fungal infections. Due to a binding site for the lipid A moiety of Lipopolysaccharide (LPS) from Gram-negative bacteria ^[18], LBP facilitates the transfer of LPS to the membrane-bound CD14-receptor ^[19]. Thereby low concentrations of LBP enhance LPS-induced cell activation and may induce inflammation at local sites of infection, whereas higher concentrations of LBP can neutralize LPS-induced activation and may prevent systemic inflammation ^[20]. LBP is constitutively present in human plasma at low concentrations (3–15 µg/ml) ^[21] and can increase up to 200 µg/ml during the acute phase response ^[22], ^[23]. There are several studies investigating the role of LBP in outcome prediction in critically ill patients with sepsis and infection, however, in part with inconsistent result ^[15], ^[22], ^[23], ^[24]. C-reactive protein (CRP) is a well established parameter to detect local infection and has also been used for many years in monitoring the inflammatory response to sepsis ^[15], ^[22], ^[24].

In the present study we compared the time course of LBP and CRP plasma levels in survivors and nonsurvivors during the first 14 days of postoperative sepsis and examined the performance of both markers regarding outcome prediction. Moreover, we studied whether LBP can provide useful information in addition to CRP and determined the correlation of LBP and CRP and their concordance concerning day-to-day changes of plasma concentrations.

Sixty consecutive patients meeting the criteria of sepsis, severe sepsis or septic shock after abdominal or thoracic surgery and admitted to the interdisciplinary ICU were screened for eligibility. One patient was excluded from analysis due to death not related to sepsis (pulmonary embolism). The demographics and clinical characteristics of the fifty-nine patients that were included in the per-protocol analysis are shown in Table 1. There were 40 (68%) survivors (group S) and 19 (32%) nonsurvivors of sepsis (group NS) 28 days after onset of sepsis (day 1) (Figure 1). The two groups had similar demographic and clinical characteristics including site of infection, surgical intervention and identified microorganisms. Both in S and NS, Gram-negative bacteria were identified in less than 50% of the patients. Likewise, severity of illness at onset of sepsis, as judged by the APACHE II score, was similar both in sepsis survivors and nonsurvivors. For the entire population studied, the APACHE II score at day 1 was 22.5±6.3 (mean ± SD) due to a high percentage (95%) of patients with severe sepsis and septic shock on admission. Within 3 days after admission, all patients enrolled were intubated, mechanically ventilated and received catecholamines (dobutamine, noradrenalin) to keep the sytemic blood pressure above 90 mmHg., and thus fulfilled the criteria of septic shock. All deaths were due to multiple organ failure.

The present study demonstrates that LBP and CRP plasma concentrations have a similar time course during the first 14 days of postoperative sepsis: First an increase with a maximum around the first or second day after onset of sepsis followed by a decrease in the first week and a re-increase in the second week. Plasma concentrations of both LBP and CRP in nonsurvivors did not significantly differ from survivors and were rather lower than those in survivors during the first five days of sepsis. Similar findings have been reported on LBP by other investigators, showing either no difference between S and NS of severe sepsis ^[15], ^[24] and critically ill patients ^[22], respectively, or significantly higher LBP levels in survivors than in nonsurvivors of severe sepsis at study entry ^[23]. In a mixed study population with SIRS, sepsis and severe sepsis, the maximum LBP concentration during the first 3 days in the ICU was found to moderately discriminate between S and NS ^[24]. However, as LBP plasma levels have been shown to correlate with sepsis severity ^[12], ^[22], differences in LBP plasma concentrations between S and NS are to be expected in inhomogeneous populations with different levels of severity. Therefore, our finding that LBP and CRP concentrations in nonsurvivors increased above that of survivors only from day 7 to day 14, but not in the first three days, may rather be due to the relatively good homogeneity of our study group regarding sepsis severity at study entry.

In the second week of sepsis, we found a marked re-increase of LBP and CRP in the nonsurviving group most likely as part of an inflammatory response to a recurrent or unresolved infection. Surprisingly, we also observed a moderate, but noticeable rise in LBP and CRP plasma levels from day 7 to day 14 in patients of the surviving group without any evidence for a new or recurrent infection. That is, we neither observed clinical and radiological signs nor significant changes in body temperature (data not shown) or white blood count during the second week in group S. We therefore assume that in these patients LBP and CRP unspecifically increased due to stressful maneuvers imposed upon them in the ICU (i.e. end of analgosedation, 9 and extubation, vigorous physiotherapy etc.). Therefore, a moderate re-increase in LBP or CRP plasma concentration in the second week of sepsis is no reliable indicator of a recurrent infection in this setting of postoperative sepsis.

Due to a broad overlap between S and NS both in LBP and CRP plasma concentrations during the first 5 days of sepsis, their predictive accuracy for mortality (at day 28) as assessed by ROC-analysis was very poor (AUC<.55). Similar findings were reported for LBP (AUC = .53) and CRP (AUC = .56) plasma concentrations on ICU admission in a heterogeneous population of postoperative patients with SIRS, sepsis or septic shock ^[24]. Likewise, Prucha et al ^[22] found no significant difference in LBP concentrations between survivors and nonsurvivors at study entry in a mixed group of critically-ill patients. In a more recent study on patients with severe sepsis, Villar et al. confirmed this result for LBP at study entry (AUC: 0.57, CI: 0.52–0.71, p = .173). However, 48 hours later, LBP plasma concentration was found a better outcome predictor (AUC: 0.71, CI: 0.61–0.80, p<.0001) ^[15]. Although the population of that study seems to be quite comparable to our study regarding sepsis severity (mean APACHE II score 23.2 vs. 22.5) and mortality (40% vs. 32%), there may be two relevant differences in the populations that would explain why we could not reproduce Villar's finding that LBP at 48 h is a good predictor of outcome: First, we studied only postoperative septic patients, whereas Villar's study seemed to include medical patients with sepsis and pneumonia as well. As shown for CRP and other acute phase proteins ^[26], postoperative LBP plasma concentrations could be unspecifically elevated due to surgical trauma. In the present study, therefore, possible differences between S and NS in the LBP levels 48 h after onset of sepsis might have been blurred by an unspecific response to surgery.

Second, in contrast to our study, where only 2 (3%) of the patients had Adult Respiratory Distress Syndrome (ARDS), 55% of Villar's patient population suffered from ARDS ^[15]. Those ARDS patients as well as the nonsurvivors were reported to have the highest values of LBP, whereas patients who survived and those who did not develop ARDS had significantly lower LBP levels at 48 h ^[15]. As previously has been shown, LBP can be produced by epithelial cells of the lungs under pathophysiological conditions found in ARDS in response to inflammatory mediators (IL-1, IL-6 and TNF alpha) ^[16]. Therefore, in Villar's study, the difference in LBP levels between NS and S may rather be attributed to a higher percentage of ARDS patients in the nonsurviving group and not merely to sepsis.

Our finding that CRP plasma levels either did not (d1–d5) or only poorly (AUC = 0.63 at d7) discriminate S from NS during the first week of sepsis, is in good agreement with the studies of Villar, Sakr and other previous investigations ^[10], ^[15], ^[24]. Furthermore, for the entire population studied and across all observation time points, we found a strong correlation between the quartile numbers LBP and CRP plasma concentrations were assigned to (r² = .95, p = .02). This result is not unexpected, since CRP and LBP are both acute phase proteins, induced by the same inflammatory mediators, and correlate significantly well with each other (r = .71, p<.001) in an experimental LPS-inhalation study in healthy volunteers ^[27]. A strong correlation (r = 0.84, p<.0001) between LBP and CRP was also found by Gaini et al. in patients with suspected community-acquired infection and sepsis ^[12] and in surgical ICU patients with sepsis syndromes (r² = .36, p<.001) ^[24]. Likewise, a significant, but less strong correlation (r = .54, p = .002) between LBP and CRP was described in patients with sepsis and septic shock ^[22].

Although the liver is the main source of both acute phase proteins, under certain pathophysiological conditions, CRP can also be produced by epithelial cells of the kidneys ^[28] as well as LBP can be released by epithelial cells of the intestine and the lungs ^[16]. Therefore, the degree of correlation between CRP and LBP may vary, dependent on the individual contribution of extrahepatic sources according to the site of infection.

More often than the absolute plasma concentration of a biomarker, its time course is used to judge the resolution of sepsis. We therefore questioned, whether a change in LBP plasma level from one observation day to the next occurs in parallel with CRP. Our results show that across all time points and all patients studied, more than three out of four changes in the LBP concentration from one to the next observation day were accompanied with a simultaneous change in the CRP level in the same direction. Although LBP is known to significantly increase in plasma approximately 12 hrs ahead of CRP in response to inflammatory stimuli ^[29], the potential of LBP to get some earlier information regarding the development of sepsis, does not really provide an advantage in a clinical setting where biomarkers are determined in 24-hour intervals. Therefore, our finding does not support the notion that assessment of LBP has significant advantage over CRP in this context. LBP and CRP plasma concentrations, assessed in intervals of 24 hours and more, rather show a very similar kinetics in the course of sepsis. Given the lower costs and the long-term clinical experience with CRP measurements, we do not consider LBP as a useful biomarker for the monitoring of sepsis in patients post surgery.

Our study has several strengths. First, according to Randolph et al. ^[30], our study population was a representative sample of critically ill adult patients and sufficiently homogenous with respect to prognostic outcome, since all patients had sepsis after major surgery and were studied from day 1 of sepsis. Second, the unbiased and well-defined endpoint used was mortality directly related to multiple organ dysfunction due to severe sepsis within the first 28 days.

However, some limitations also merit consideration: The prognostic accuracy of LBP and CRP may have been negatively influenced in our study population, because LBP and CRP are frequently unspecifically elevated after major surgery to levels comparable to those seen in sepsis. Therefore, in septic patients not subjected to surgical trauma, e.g. in medical patients, LBP and CRP may provide a better prognostic accuracy. In addition, although statistically controlled, we cannot completely exclude small sample size effects in our study population. Moreover, LBP and CRP may be clinically useful in critically ill neonates and children, where LBP has a high diagnostic accuracy to differentiate sepsis from non-infectious SIRS ^[14], ^[31].

In conclusion, our study demonstrates that LBP and CRP plasma concentrations are well correlated with each other and change concordantly in the course of sepsis. Furthermore, LBP and CRP plasma levels have quite the same predictive value and are neither suitable to monitor resolution of sepsis nor sufficiently reliable to detect a recurrent infection during the first 14 days of sepsis in adult postoperative patients.

Competing Interests: The authors have declared that no competing interests exist.

Funding: The study was supported by the Department of Anesthesiology, University of Erlangen-Nürnberg. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

The authors would like to thank Harald Ihmsen, PhD, for his help and excellent advice.