Introduction
Acute respiratory distress syndrome (ARDS) is a disorder historically consisting of hypoxemia accompanied by bilateral pulmonary infiltrates and has been identified as a major contributor to mortality in the pediatric intensive care unit (PICU) population [
1,
2]. ARDS involves disruption of the integrity of both the pulmonary endothelium and alveolar epithelium, leading to the development of non-cardiogenic pulmonary edema [
3]. While both epithelial and endothelial surfaces are involved in the pathophysiology of the disease, the location of initial insult has classically been used to classify the disease as direct or indirect. Direct lung injury results primarily from injury to the pulmonary epithelium, such as pneumonia or aspiration. In contrast, indirect lung injury develops primarily through damage to the pulmonary endothelium from extrapulmonary sources of systemic inflammation such as sepsis, trauma, or transfusions [
4]. Though there is evidence of endothelial injury in patients with ARDS of any etiology, studies in adult patients with ARDS have demonstrated evidence of increased endothelial injury among those with indirect lung injury as compared to those with direct lung injury [
5‐
7]. In addition, endothelial dysfunction is associated with the development of multiple organ dysfunction syndrome (MODS) [
8], which is a major mediator of mortality in ARDS [
9]. As such, assessment of circulating measures of endothelial injury may improve our understanding of the pathogenesis of ARDS in children, provide a method to risk-stratify children with ARDS, and provide supportive evidence to guide development of future therapeutic interventions.
Thrombomodulin is a transmembrane protein present on all endothelial surfaces and is highly expressed in pulmonary alveolar capillaries [
10]. It facilitates the thrombin-mediated conversion of protein C to activated protein C and has roles in coagulation, fibrinolysis, and inflammation [
11,
12]. During normal health, thrombomodulin (TM) sheds from the endothelial surface, and soluble thrombomodulin (sTM) circulates at low levels [
12]. These circulating fragments have markedly attenuated anticoagulant activity compared with cellular TM or recombinant TM [
13,
14]. In the presence of inflammation, neutrophil proteases increase the release of sTM from the cell surface and thereby increase circulating levels of sTM [
15]. Studies in children with meningococcal sepsis have documented this effect by demonstrating both depletion of endothelial thrombomodulin and simultaneously increased plasma sTM as compared to controls [
16]. sTM levels are similarly elevated in sepsis, disseminated intravascular coagulation (DIC), vasculitis, venous thrombosis, and trauma [
17‐
19]. Experimental and clinical studies have identified sTM as a marker of generalized endothelial injury [
20,
21].
sTM has been found to be elevated in the plasma and alveolar edema fluid of adults with ARDS [
22], and increased plasma sTM levels are correlated with worse clinical outcomes in adults with ARDS [
23‐
25]. The role of endothelial damage as measured by sTM in children with ARDS and its relative importance in direct versus indirect lung injury are unclear. We hypothesized that sTM levels would be associated with mortality and organ failure in a prospectively studied pediatric population with ARDS and that the association might differ based on the mechanism of lung injury.
Methods
Design and patient population
Data were collected between 2008 and 2014 as part of a prospective, multicenter observational cohort of pediatric acute lung injury. Patients were screened for eligibility throughout their stay in five PICUs in California and Wisconsin. Eligibility criteria included acute onset of respiratory symptoms requiring non-invasive or invasive positive pressure support. Patients met the American-European Consensus Conference criteria (referred to in the text as ARDS) [
4]. Briefly, eligible patients had an arterial partial pressure of oxygen (PaO2 mmHg) to the fraction of inspired oxygen (FiO2) ratio (P/F ratio) <300 mmHg and had new bilateral infiltrates on chest radiography as judged by site investigators. Patients could also become eligible based on pulse oximetry (SpO2) and an SpO2/FiO2 ratio <253 when PaO2 values were not available [
26]. Patients were excluded if they were <30 days of age, <36 weeks corrected gestational age, >18 years of age, had a documented limitation of care order, were wards of the state at the time of screening, or had been enrolled in the cohort previously.
Ethics, consent and permissions
The institutional review board of the University of California, San Francisco, CA, Oakland Children’s Hospital and Research Center, Oakland, CA, Valley Children’s Hospital, Madera, CA, Children’s Hospital Los Angeles, Los Angeles, CA, and the University of Wisconsin, Madison, WI reviewed and approved the collection of clinical data and biological samples. Informed consent was obtained from patients’ parents or legal guardians.
Data and sample collection
Demographics, and pre-existing medical conditions, as documented in the medical record, were collected at enrollment. The primary lung injury risk factor was recorded as determined clinically by site investigators. The lung injury group (direct vs. indirect) was then assigned as previously described [
4]. After enrollment, daily data including vital signs, respiratory therapies, and laboratory results were recorded. Pediatric risk of mortality (PRISM) III scores were calculated from these data [
27]. Vasopressor use was recorded for any dose of one or more of the following: dopamine, epinephrine, norepinephrine, phenylephrine, dobutamine, or milrinone. Patients were followed through hospital discharge. Plasma samples for sTM assays were collected in EDTA tubes within 24 hours of meeting study eligibility criteria. Plasma samples were obtained only from indwelling catheters or during scheduled laboratory collections.
The primary outcome was death prior to intensive care unit (ICU) discharge (ICU mortality). The secondary outcome was the pediatric logistic organ dysfunction (PELOD) score, which is a validated outcome measure of organ dysfunction in critically ill children [
28]. The single worst value for each component of the six-organ systems from days 1-7, 14, 21, and 28 after study enrollment were used in the calculation of the PELOD score. The number of failing organ systems was defined as the number of organ systems with a PELOD subscore ≥1 [
28].
Soluble thrombomodulin assay
sTM was measured in plasma samples using two-antibody sandwich enzyme-linked immunosorbent assays (Asserachrom Thrombomodulin assay, Diagnostica Stago, Parsippany, NJ, USA). Samples were assayed in duplicate according to the manufacturer’s protocol and the mean value used for analysis. Samples were used for analysis if the variance between duplicate sTM assays was ≤15 %.
Statistical analysis
Normally distributed continuous data are reported as mean and standard deviation, and were compared by
t test. Non-normally distributed, skewed data are reported as median and interquartile range (IQR), and were compared by the Mann–Whitney
U test. Cuzick’s non-parametric test of trend was used to assess for monotonic trends across ordered groups [
29]. Non-normally distributed, skewed variables were log-transformed for use in parametric tests and for graphical representation as indicated. Categorical variables were analyzed by the chi square (χ
2) test. We used Spearman rank correlation coefficients (rho) to assess and test relationships between variables without assuming normal distributions. We used logistic regression models to quantify the relationships between sTM and mortality and to adjust for covariates. Initially, we assessed the effect of site on the analysis by including site as a random intercept in a mixed effects logistic analysis. The estimated variance of the random intercept was extremely small, consistent with there being no site effect. We therefore used standard logistic regression models for the remainder of the analysis. In view of our sample size we used bivariable models to adjust for one covariate at a time. The Pearson chi square test was used to assess the goodness of fit (GOF) of the data to a logistic model.
We used linear regression analysis to quantify the relationship between sTM and PELOD and to adjust for clinically significant covariates. We adjusted for clinically significant covariates such as age, PRISM III, P/F ratio, and the presence of acute kidney injury (AKI). These covariates were chosen because sTM levels have been reported to vary with age [
30], to avoid confounding by the severity of illness and initial lung injury, and AKI was included because sTM is known to be primarily excreted via the kidneys [
14]. Due to the small numbers of deaths, the adjusted logistic models for mortality each only included a single potential confounding variable. Patients were deemed to have AKI if one or more of the following criteria were present: urine output <0.5 mL/kg/h, estimated glomerular filtration rate <50 mL/min/1.73 m
2 body surface area, or receipt of hemodialysis [
31]. Receiver operator characteristic curve analysis was used to assess the discrimination of sTM for mortality. Statistical analysis was carried out using Stata 13 (Stata Corporation, College Station, TX, USA).
Discussion
In this study we found that (1) elevated plasma sTM levels within 24 hours of diagnosis in children with ARDS are associated with organ dysfunction, (2) elevated plasma sTM is associated with higher mortality among patients with indirect lung injury, and (3) sTM levels are elevated in indirect lung injury when compared with direct lung injury. These findings emphasize the importance of endothelial injury among pediatric patients with ARDS, especially in patients with an indirect mechanism of lung injury.
The findings are consistent with the need for prolonged respiratory support [
24] and increased mortality with elevated plasma sTM among adult patients from the ARDS network studies [
25]. The findings are also consistent with studies that reported elevated plasma sTM in pediatric sepsis [
16,
32] and associations between plasma sTM and sepsis-related mortality and organ dysfunction in adults [
33]. However, the finding in our cohort that organ dysfunction scores also correlate with plasma sTM levels in direct lung injury implies that the relationship is not restricted to septic patients.
Elevated plasma sTM early in the course of the ARDS disease process provides a potential surrogate measurement for the degree of endothelial damage. This is supported by the location of thrombomodulin on the endothelial cell membrane and experimental evidence demonstrating that it is released into the circulation under conditions of inflammation [
20]. In addition, plasma sTM levels correlate with the concentration of circulating endothelial cells, which is a metric of endothelial damage [
21]. Previous studies of adult patients have also found evidence that endothelial activation and injury are important in the pathogenesis of ARDS. Von willebrand factor (VWF), another potential marker of endothelial dysfunction produced by endothelial cells and platelets, is elevated in patients with non-pulmonary sepsis who go on to develop ARDS [
34]. VWF early in the course of ARDS has also been associated with mortality and increased rates of organ failure in adults and a small study in children [
35,
36]. In another adult study increased plasma angiopoietin 2, a marker of endothelial dysfunction and vascular permeability, was associated with increased mortality and fewer ventilator-free-days [
37]. The outcome of our study is consistent with these findings and helps extend the importance of endothelial injury and dysfunction in ARDS to pediatric populations.
Another possible mechanism for the observed association between elevated plasma sTM and adverse clinical outcomes is through increased intravascular thrombosis, leading to impaired microcirculation and the development of organ dysfunction. Decreased intact cellular thrombomodulin in the setting of inflammation [
16] and the observation of relatively decreased anticoagulant activity of its soluble form [
13] suggest that patients with elevated plasma sTM are at increased risk of organ injury. Studies of ARDS in animal models and human subjects have suggested that pulmonary coagulopathy plays a significant role in the pathobiology of this disease and its associated comorbidities [
38]. Relative deficiencies in proteins such as protein C, antithrombin, tissue factor pathway inhibitor, plasminogen activator inhibitor 1, and thrombomodulin have been implicated in this disruption of the normal balance of procoagulant and anticoagulant forces [
38,
39]. The nature of this mechanism suggests that soluble thrombomodulin may be an important mediator of disease in ARDS and an early marker of disease severity.
Despite biological plausibility and promising results in animal models of sepsis and acute lung injury, repletion of natural anticoagulant proteins has thus far shown limited success in human trials [
38]. However, none of these trials used biological markers to target these therapies to specific patient populations. Clinical trials have investigated the use of recombinant soluble thrombomodulin therapy in sepsis, DIC, and ARDS [
40,
41], though no conclusive pediatric data are available as of yet. These trials did not report pre-treatment sTM levels or use plasma sTM as a basis for targeted therapy with recombinant TM. The results suggest that pediatric patients with ARDS, especially sicker patients with elevated plasma levels of sTM, may be potential candidates for future studies of recombinant thrombomodulin replacement therapy.
The primary strengths of this study are the relatively large size of our pediatric ARDS cohort and the prospective collection of biological samples, which is unusual in this population. The absolute numbers of patients and outcome events also limited our analysis, especially the ability to adjust for multiple potential confounders within a single logistic regression model. We therefore used multiple bivariable models to adjust for one potential confounder at a time. After adjustment for the initial severity of illness by PRISM III the association between sTM and mortality was no longer statistically significant in our logistic model. The magnitude of the association was only slightly attenuated, and the loss of significance may have been related to inadequate sample size. However, our population is reflective of the relatively low frequency of ARDS and decreasing mortality [
2] from ARDS in children, and nonetheless we were able to identify significant association between sTM and mortality among patients with indirect lung injury. Another limitation is our inability to obtain plasma samples for analysis of the entire cohort. However, the patients without plasma samples had less organ dysfunction and lower absolute mortality, indicating that those patients were less severely ill and consistent with the fact that they did not require indwelling central lines or arterial lines from which to draw samples. This may limit applicability of our findings to more critically ill ARDS patients.
Conclusions
In summary elevated soluble thrombomodulin, when measured in children early in the course of ARDS, is associated with increased organ dysfunction and also is associated with increased odds of mortality among children with an indirect mechanism of lung injury. These findings should be validated in additional populations, but they provide evidence of the important role of the magnitude of endothelial injury in determining outcomes from pediatric ARDS. Future interventions targeted toward endothelial stabilization, repair, or functional supplementation may benefit this population.
Acknowledgements
Funds provided through NIH grants NHLBI HL085526 (K23) (Sapru), NHLBI HL114484 (R01) (Sapru), and NHLBI R37HL51856 (Matthay). Thanks to Victoria Lo, Elizabeth Colglazier, and Izabella Damm for help with initial study sample collection and processing. We would also like to thank Dr. Patrick McQuillen for his comments and manuscript review.
Competing interests
The authors report no financial ties to products used in the study, or conflicts of interest.
Authors’ contributions
BO assisted with study design, conducted data collection, statistical analysis, and drafted the manuscript. ACS performed statistical analysis and data collection. MZ assisted in data collection and analysis. MA assisted with study enrollment and sample collection, and conducted the soluble thrombomodulin assays. RK, HF, CC, and MM assisted with study design and development of the study cohort. JN assisted with statistical analysis. AS conceived of the study cohort and study design, and assisted with study coordination and drafting of the manuscript. All authors participated in critical review and revision of the manuscript, and approved the final form.