Heparin-binding protein as a biomarker of severe sepsis in the pediatric intensive care unit: A multicenter, prospective study

Background: Heparin-binding protein (HBP) is a promising candidate as a biomarker for sepsis. However, there is limited study on the use of HBP among children with sepsis in pediatric intensive care unit (PICU). The aim of this study is to assess HBP as a diagnostic and prognostic biomarker of severe sepsis in the PICU. Methods: A multicenter, prospective study was conducted among children with sepsis and severe sepsis in nine different PICUs in China from October 2019 to June 2021. Plasma levels of HBP, procalcitonin (PCT), C-reactive protein (CRP), lactate, and white blood cell (WBC) count were determined at enrollment and 72 hours after enrollment. Receiver operating characteristic curve (ROC) analysis was used to evaluate the ability of biomarkers in diagnosing severe sepsis. Multivariate logistical analysis was performed to assess the association between biomarkers and in-hospital mortality. Spearman's correlation was used to identify the relationship between HBP and other biomarkers. Results: Of 355 included patients, 132 patients were diagnosed with non-severe sepsis (referred to as sepsis), 223 patients had severe sepsis. Patients with severe sepsis had signi�cantly elevated levels of HBP compared with sepsis (median 170.5 vs. 74.1 ng/mL, P<0.001). Adding HBP to a diagnostic model with PCT and lactate could signi�cantly improve the diagnostic capability for severe sepsis (area under the curve (AUC) 0.702 vs. 0.628, p<0.001). The plasma levels of HBP correlated positively with the number of dysfunctional organs. After adjusting for confounding factors, the HBP levels at enrollment could not predict in-hospital mortality. However, declined levels of HBP at 72 hours had a signi�cant association with decreased in-hospital mortality (adjusted odds ratio (aOR) 0.242, P<0.001). The levels of HBP showed weak positive correlations with PCT, CRP, WBC, and no correlation to lactate. Conclusions: HBP at enrollment can be an independent indicator for


Background
Sepsis is a major contributor to morbidity and mortality in children globally. 1 In 2017, an estimated 48.9 million (95% CI, 38.9-62.9) incident cases of sepsis were recorded worldwide and 11.0 million (95% CI, 10.1-12.0) sepsis-related deaths were reported, representing 19.7% (95% CI,18. 2-21.4) of all global death. The percentage of sepsis-related deaths was highest in early childhood among all age groups.
There were an estimated 2.9 million (95% CI, 2.6-3.2) deaths related to sepsis worldwide among children younger than 5 years. 2 Severe sepsis is a life-threatening condition commonly treated in pediatric intensive care unit (PICU) worldwide. A meta-analysis based on 94 studies including 7561 children showed that the mortality rate in pediatric severe sepsis and septic shock was 31.7% (95% CI, 27.3%-36.4%) in developing countries, and 19.3% (95% CI, 16.4%-22.7%) in developed countries. 3 Considering the high mortality of severe sepsis early recognition and risk-strati cation could assist clinicians to manage the care of these patients more effectively and to improve outcome. 4 Laboratory tools to identify patients with severe sepsis early and to stratify the severity are in high demand. Although a multitude of biomarkers have been proposed in the eld of sepsis, so far none is considered adequate for routine clinical use. 5 Heparin binding protein (HBP), also known as azurocidin or CAP37, is an antimicrobial protein stored in the secretory vesicles and azurophilic granules of neutrophils and is released when neutrophils are activated. 6 The release of HBP has been demonstrated in various infectious diseases such as meningitis, pneumonia and urinary tract infection. [7][8][9][10] HBP also has been shown to be the link in neutrophil-derived induction of vascular leakage. 11 These characteristics make HBP a promising candidate as a biomarker for sepsis. Previous studies have successfully evaluated plasma HBP for prognosticating infection-induced organ dysfunction (OD) in sepsis and disease severity. [12][13][14][15] Most of published results regarding the powerful diagnostic and prognostic capability of HBP in adult patients with sepsis. However, the clinical value of HBP in pediatric patients has not been comprehensively assessed. Therefore, in this study, a multicenter prospective study was performed to investigate the plasma levels of HBP among children with sepsis in PICU, aiming to 1) assess the clinical validity of HBP for diagnosing severe sepsis, and predicting in-hospital mortality, and 2) to compare the properties of HBP to other sepsis biomarkers currently in use, including lactate, procalcitonin (PCT), C-reactive protein (CRP) and white blood cell (WBC).

Study design and participants
This was a prospective, multicenter study of PICU patients with sepsis, conducted at nine children's hospitals in China from October 2019 through June 2021. This study was approved by the institutional ethics committees of the participating hospitals.
Patients < 18 years of age admitted to PICUs with sepsis were recruited. Sepsis and severe sepsis were de ned according to the 2005 International Pediatric Sepsis Consensus Conference. Sepsis was de ned as con rmed or suspected invasive infection with two or more systemic in ammatory response syndrome criteria; severe sepsis was de ned as sepsis in addition to one of the following: cardiovascular dysfunction, acute respiratory distress syndrome, or two or more other organ dysfunctions (respiratory, renal, neurologic, hematologic, or hepatic). 16 Patients with neutropenia due to hematological malignancies were excluded. The nal diagnosis was made by the attending physicians who were unaware of the research results.

Sample and laboratory analysis
Blood samples were collected in BD sodium citrate vacuum tubes (1.8ml) from patients at the time of inclusion (at enrollment) and again 72 hours after the initial sample (72 hours).
Samples were processed locally at each site, centrifuged at 1000 g for 10 minutes. HBP, PCT, and CRP were measured in plasma by a uorescence dry quantitative immunoassay using the Jet-iStar 3000 analyzer (Join-Star, Hangzhou, China) according to the manufacturer's recommendations. In addition, laboratory tests including WBC count and lactate were performed as part of clinical routine at each site.

Clinical data collection
All clinical data of enrolled patients were obtained from medical records, including demographics, comorbidities, medication usage, underlying infections, organ dysfunction and outcomes.

Statistical analysis
Categorical variables were expressed as numbers (%) and compared using the chi-square test or Fisher's exact test as appropriate. Continuous variables are presented as medians (interquartile range, IQR) and were analyzed with using Mann Whitney U test.
Receiver operator characteristics (ROC) curves were drawn and the areas under the curve (AUC) were calculated to assess the diagnostic power of biomarkers and different diagnostic models. To assess the value of HBP on diagnosing severe sepsis beyond other conventional biomarkers, we established 2 diagnostic models. Model 1 was a conventional diagnostic model including PCT and lactate. These biomakers were chosen based on ROC analyses for each biomarker in the study. Model 2 was a novel model which additionally add HBP to the model 1. The diagnostic models were based on the predicted probabilities of a multivariate logistic regression model. The AUCs for these were compared using the DeLong's test. Youden's index was used to choose the best cut-off points for assessment of sensitivity and speci city.
Jonckheere-Tepstra trend test was used to verify whether the levels of HBP increased with the number of dysfunctional organs.
Multivariate logistic regression analysis was done to evaluate the association of HBP and other biomarkers (levels at enrollment; dynamic changes at 72 hours) with in-hospital mortality among children with sepsis. We adjusted for age, sex and the presence of comorbidity. The results were described as the adjusted odds ratio (aOR) and 95% con dence interval (CI).
Spearman non-parametric correlation was used for calculating the correlations between HBP and other biomarkers.
All of the tests were two tailed, and a value of P < 0.05 represented statistical signi cance. Statistical analyses were conducted in SPSS version 26.0 software (IBM, New York, USA) or MedCalc version 20.1 software (MedCalc, Mariakerke, Belgium).

Results
Characteristics of the patients Table 1 displays the baseline clinical characteristics of the patients included in the study. A total of 355 PICU patients with sepsis were enrolled in the study. According to the 2005 International Pediatric Sepsis Consensus, the research subjects were divided into non-severe sepsis group (referred to as sepsis group) (n=132) and severe sepsis group (n=223). The overall median age was 15 months (IRQ 5 to 55) and 58.6% were male. Patients with severe sepsis were older than patients with sepsis.
One hundred fty-one (42.5%) patients had at least one comorbidity. The most common comorbidity was immunological diseases (11.8%), followed by cardiovascular diseases (8.2%), gastrointestinal diseases (5.4%) and metabolic diseases (3.7%). There was no signi cant difference in the proportion of comorbidities among groups. The respiratory tract infection was the predominant source of infection in all groups, followed by gastrointestinal tract, central nervous system and genitourinary infections. Respiratory tract infections had a higher prevalence in the severe sepsis group. Of all patients, 290 (81.7%) had at least one organ dysfunction with respiratory (59.2%) and cardiovascular (40.0%) dysfunctions the most common.
The in-hospital mortality of the entire cohort was 20.3%, while the median length of PICU stay were 9 days (IRQ 6 to 15). Mortality among patients in severe sepsis group was signi cantly higher compared to sepsis group. There was no signi cant difference in the length of PICU stay between groups.
Higher HBP levels in patients with severe sepsis At the time of inclusion, the concentration of HBP in the severe sepsis group (median 170.5 ng/mL, IQR 46.1-300) was signi cantly greater than that in the sepsis group (median 74.1 ng/mL, IQR 24.5-175.5).
PCT and lactate levels were also signi cantly higher (P<0.01) but there was no signi cant difference in CRP (P=0.892) or WBC levels (P=0.088) in the severe sepsis group compared to sepsis group ( Figure 1).
HBP signi cantly improves diagnostic value ROC analyses were performed for HBP and other biomarkers to discriminate between severe sepsis and sepsis. The AUC for HBP, PCT and lactate was 0.640 (95% CI: 0.582-0.698), 0.623 (95% CI: 0.564-0.683), and 0.589 (95% CI: 0.526-0.652), respectively ( Table 2). The optimal cutoff level was 193.6 ng/ml for HBP, with sensitivity of 0.48 and speci city of 0.81. Although the AUC for HBP was higher than PCT and lactate, there was no signi cant difference when using DeLong's test (Table 2) (Figure 2A).
We next explored whether inclusion of HBP could improve and allow better recognition of patients with severe sepsis. The diagnostic value was assessed using 2 models (Model 1: PCT and lactate; model 2: model 1 + HBP). The AUC for model 2 was 0.702 (95% CI: 0.647-0.752), which was signi cantly higher than model 1 (AUC: 0.628; 95% CI: 0.571-0.682) ( Table 2) ( Figure 2B). Thus, the inclusion of HBP to the model with PCT and lactate levels signi cantly improved the diagnostic value for severe sepsis.
HBP levels are related to the severity of the disease The severity of the sepsis was strati ed based on number of dysfunctional organs. Median HBP level at the inclusion signi cantly increased along with the number of dysfunctional organs with a monotonic trend, as demonstrated with a Jonckheere-Terpstra test ( Figure 3A). Statistically signi cant increasing trends were also observed in the level of PCT and Lactate ( Figure 3B, C), while the level of WBC decreased as the number of dysfunctional organs increased ( Figure 3D). There was no signi cant association between CRP level and the number of dysfunctional organs ( Figure 3E).

Discussion
Early diagnosis and evaluation of severe sepsis are crucial for physician to modify treatments and associated with decreased in-hospital mortality. 17 In this multicenter prospective study, the utility of HBP was assessed among children with sepsis in PICUs. We found HBP to be a promising biomarker that distinguished severe sepsis from sepsis and combination of HBP with PCT and lactate could signi cantly improve the diagnostic value for severe sepsis. Moreover, the dynamic changes of HBP at 72 hours after enrollment could predict in-hospital mortality among children with sepsis in PICU.
HBP acts as an ampli er of in ammatory responses and induces capillary leakage, both of which are highly dysregulated in severe sepsis. 6 As expected, we observed that the levels of HBP at enrollment were signi cantly higher in severe sepsis group (median 170.5 ng/mL) compared with sepsis group (median 74.1 ng/mL). Our results are consistent with previous researches in adult patients. Linder et al. reported that HBP levels were signi cantly higher in the severe sepsis group (median 28 ng/mL) as compared to the patients with non-septic critical condition (median 14.5 ng/mL) in ICU. 14 Similar results were also obtained among adult patients with sepsis under Sepsis-3 criteria (sepsis de ned as organ dysfunction caused by infection). 18-20 In addition, our study showed that the HBP levels at enrollment signi cantly increased along with the number of dysfunctional organs, implying that HBP is a marker of disease severity. This is in keeping with a study conducted in adult Emergency Department. 13 Taken together, those results further prove the key role of HBP in the pathophysiology of organ dysfunction in sepsis, not only in adults but also in children. Of note, the median values of HBP in children with severe sepsis were signi cantly higher than in adults reported in previous studies. 14, 19-21 One plausible explanation for these differences is that the activation of peripheral blood neutrophils might be varied among children and adults. Differences in baseline characteristics such as the presence of comorbidity or receipt of steroids, may also explain such differences. Larger cohort studies are needed to further verify those results.
The diagnostic capability of HBP as a biomarker in severe sepsis has been evaluated in previous cohort studies among adults. Linder et al. reported that a cut-off level for HBP of ≥ 15 ng/mL showed a sensitivity of 87.1% and a speci city of 95.1% in diagnosing severe sepsis from control group (sepsis, infection without SIRS, and SIRS without infection). 21 Zhou et al. revealed that a cut-off value for HBP of ≥ 28.1 ng/mL gave a sensitivity of 84.9% and a speci city of 78.3% in diagnosing sepsis (sepsis-3 criteria) from infected people. 20 The diagnostic capability of HBP was higher than other tested biomarkers (e.g. PCT, CRP and lactate) in both studies. In the present study, the results of the ROC curve analysis also showed that HBP had the highest AUC, followed by PCT and lactate. However, there was no signi cant difference of AUC between HBP and PCT or lactate when using Delong's test, which means the diagnostic capability of HBP was not signi cantly superior to PCT and lactate in distinguishing severe sepsis from sepsis. Whereas, the diagnostic capability was signi cantly improved when adding HBP to the diagnostic model with PCT and lactate (AUC 0.702 vs. 0.628, P < 0.001). We also found that there was little correlation of HBP with other biomarkers, indicating that HBP provides separate information as a biomarker. These observations suggested that HBP levels were associated with severe sepsis independent of PCT and lactate levels, and the combination of HBP, PCT and lactate to diagnose children with severe sepsis was considered as a superior approach in clinical practice.
We found that initial plasma HBP at enrollment was not associated with in-hospital mortality after adjusting for age, sex and the presence of comorbidity. The results were concordant with prior studies of HBP in sepsis, which showed that HBP could not predict 28-day mortality in the emergency department and the plasma levels of HBP did not correlate with survival in non-septic and septic patients with shock. 19,22 However, an important new nding in our study was that the declined level of HBP at 72 hours had a signi cant association with decreased in-hospital mortality and the prognostic value of HBP dynamic changes was better than the other biomarkers. These ndings suggested that serial plasma HBP measurements provided useful prognostic information in children with sepsis and the dynamic of HBP could potentially be used to inform medical decision to improve patient outcomes in PICU.
Our study had several strengths. First, this was the rst large prospective, multicenter study evaluating the role of HBP as a biomarker in the diagnosis and prognosis of severe sepsis in children. Second, the patients included in this study involve a broad range of diagnoses and comorbidities in nine PICUs in China, which provided more applicable results for children with sepsis in PICU. Third, the plasma levels of HBP, PCT and CRP were simultaneously detected in the same plasma sample, which made the results more comparable. Fourth, with all patients reviewed by at least two PICU physicians according to the 2005 International Pediatric Sepsis Consensus, the diagnosis of sepsis was accurately determined.
The study also had some limitations. First, this study included only pediatric patients who were diagnosed with sepsis and critically ill in PICU. Thus, the results may not be generalizable to patients with sepsis in emerging department or general ward. Second, there were a lot of HBP levels exceeding the detection limit (300 ng/mL), especially in the severe sepsis group, which might decrease the statistical power. Third, critically ill patients without sepsis were not enrolled as control in this study. Chew et al.
found elevated HBP levels in some patients with noninfectious shock. 22 Further investigations are needed to better address to what extent HBP is speci cally associated with sepsis in children.

Conclusions
Our study reveals that the initial plasma HBP level at enrollment could be a useful biomarker to diagnose severe sepsis and was related to the severity of disease in PICU. The declined level of HBP at 72 hours had a statistically signi cant association with decreased patient mortality. The levels of HBP showed weak positive correlations with PCT, CRP, WBC, and no correlation to lactate. Further studies are needed to validate these ndings in patients with sepsis in PICU.    Abbreviations: ROC, receiver operator characteristics; HBP, heparin binding protein; PCT, procalcitonin; WBC, white blood cell; CRP, C-reactive protein.  Multivariable logistic regression analysis evaluating the association between biomarkers and in-hospital mortality among patients with sepsis. Confounding factors including age, sex and the presence of comorbidity were adjusted in the model. aOR CIs signi cantly associated with mortality (P < 0.05) are shown in red when aOR > 1 and in blue when aOR < 1. aOR, adjusted odds ratio; CI, con dence interval; HBP, heparin binding protein; PCT, procalcitonin; WBC, white blood cell; CRP, C-reactive protein.