Trauma is the leading cause of death in the United States for people younger than 45 years (1) and incurs substantial short- and long-term disability (1). Outcomes of injury following trauma depend on the extent of injury, the host’s attendant inflammatory and pathophysiologic responses, and timely care (2). A core question that has not yet been addressed is a basic one: “Does systemic inflammation increase as a function of injury severity, or does the character of this dynamic inflammatory response differ significantly as injury becomes greater?” Answering this question in a clinical setting should take into account that trauma patients differ based on type of injury, location, and additional factors such as the individual’s sex, the length of the interval between injury and surgery, the presence of shock at time of presentation, and the presence of cranial injuries (3). Because these different presentations of trauma along with the individual-specific demographic factors can in turn impact the inflammatory response differentially (4), an important starting point is the choice of injury severity assessment, which must be able to compare disparate types of trauma meaningfully (5,6).
The need to improve the quality of trauma care—rather than a focus on the question of how injury severity affects physiology at a basic level—has led researchers over the past few decades to develop tools that allow physicians to predict the outcomes in polytrauma patients (6), including many scoring systems (7). Among these injury severity scoring systems, the Injury Severity Score (ISS) (8), Revised Trauma Score (9), and Trauma Injury Severity Score (10) have gained popularity and are widely used. Despite the known limitation in the ISS score, it is still the one most commonly used to assess the extent of anatomic trauma injuries despite their cause (11) in most institutions, likely because of the ease in calculation. Although the ISS was developed in order to predict mortality (8), this scoring system has also been used to predict other factors such as clinical outcomes (12) and nosocomial infections (NIs) postinjury (13). Compared with the other injury severity scores, the ISS comes closest to quantifying the degree of externally driven injury (not the host’s own physiological responses to that injury).
Systemic inflammation ensues in response to injury. To date, the focus of studies on systemic inflammation as a function of increasing injury severity has been on defining individual mediators that are elevated, typically in an attempt to help to better stratify patient outcomes (14). However, it is now increasingly clear that in order to gain new insights into the mechanisms of postinjury inflammation, as well as to define new diagnostic and therapeutic modalities, the inflammatory response must be viewed as a system of dynamic, interconnected processes (15, 16). Herein, we hypothesized that differential dynamic trajectories and networks of systemic inflammation are set in motion by injuries of different severity, in a manner that corresponds to increasingly adverse outcomes as a function of injury severity. Accordingly, we sought to examine the impact of injury severity on the systemic inflammatory response—both in magnitude and interconnectivity—and to define the associated downstream clinical outcomes.
We analyzed data retrospectively from a large cohort of blunt trauma survivors (472 patients) studied over an 8-year period (17). Frequent sampling (three samples within the first 24 h after injury and daily up to day 7 after injury) allowed substantial precision to characterize the early and late inflammatory response posttrauma. To reduce the effects of confounding variables, we derived stringently matched subcohorts of mildly, moderately, and severely injured patients that still reflected the primary demographic and injury characteristics of the original large cohort. Utilizing emerging computational methods, we defined dynamic inflammation networks in these subcohorts. Persistent elevation of interleukin 6 (IL-6) after injury is associated with worse outcomes in the severely injured cohort. Our analyses suggest that different injury severities lead to injury-specific dynamic patterns of inflammation centered around differential impact of chemokines (monocyte chemotactic protein 1 [MCP-1]/CCL2, monokine inducible by interferon γ [MIG]/CXCL9, and IP-10/CXCL10) on the key cytokine IL-6.
MATERIALS AND METHODS
Patient enrollment and sampling
All human sampling was done following approval by the institutional review board of the University of Pittsburgh, and informed consent was obtained from each patient or next of kin as per the institutional review board regulations. Patients eligible for enrollment in the study were at least 18 years of age, were admitted to the intensive care unit (ICU) after being resuscitated, and per treating physician were expected to live more than 24 h. Reasons for ICU admission included (1) hemodynamic instability requiring vasoactive drips; (2) fluid resuscitation or blood replacement needing invasive monitoring; (3) severe hypoxia necessitating complex ventilator management; (4) elderly with preexisting coronary, pulmonary, or peripheral vascular disease; and (5) patients status post damage control surgery with open abdomen or chest.
Reasons for ineligibility were isolated head injury, pregnancy, and penetrating trauma. Laboratory results and other basic demographic data were recorded in the database via direct interface with electronic medical record. Three plasma samples, starting with the initial blood draw upon arrival, were assayed within the first 24 h following trauma and then from days 1 to 7 after injury. The blood samples were centrifuged, and plasma aliquots were stored in cryoprecipitate tubes at −80°C for subsequent analysis of inflammatory mediators (see below).
Study design and selection criteria
This was a retrospective case-control study, the salient characteristics of which were described recently (17). In brief, clinical data from 472 blunt trauma survivors (330 males and 142 females, aged 48.4 ± 0.9 years, ISS 19.6 ± 0.5) admitted to the emergency department of the University of Pittsburgh Medical Center–Presbyterian (a level I trauma center) were analyzed as a function of injury severity. In an attempt to reduce the impact of any confounding factors present in the general cohort and similar to recent studies from our group (17–19), we used more stringent filtering criteria for the 472 patients by excluding those patients who had fewer than three blood samples in the first 24 h after injury, patients with documented alcohol intoxication, and patients for whom data were incomplete. Furthermore, we matched patients according to age, sex, and mechanism of injury (motor vehicle collision [MVC], falls, and motorcycles). Our final study cohort consisted of 49 mildly injured patients, 49 moderately injured patients, and 49 severely injured patients who were as well matched as possible. The overall demographics, mechanisms of injury, and clinical data of the 147 trauma patients are shown in Table 1. Importantly, this subcohort represents the age, the sex ratio, and mechanisms of injury ratios of the general cohort.
Clinical data collection
Clinical data, including ISS (see below for calculation), Abbreviated Injury Scale (AIS), ICU length of stay (LOS), hospital LOS, days on mechanical ventilator, Marshall Multiple Organ Dysfunction (MOD) score, prevalence of prehospital hypotension, prevalence of NIs, surgical interventions, biochemical parameters (hemoglobin [Hgb], hematocrit [Hct], platelet count, white blood cell [WBC] count, lactate, base deficit [BD]), heart rate (HR), respiratory rate (RR), temperature (T), and patient dispositions, were collected from hospital inpatient electronic trauma registry database. Laboratory results and other basic demographic data were obtained from the electronic medical record.
Injury Severity Score calculation
Injury Severity Score (8,20) and AIS-05 scores (according to the updated 2005 injury code) (21) were calculated for each patient by a single trauma surgeon after attending radiology evaluations were finalized. The ISS is based on an anatomical scoring system that provides an overall score for patients with multiple injuries (8). Each injury is assigned an AIS score, allocated to one of six body regions: head and neck, face, chest, abdomen, extremities (including pelvis), and external. Only the highest AIS-05 score in each body region is used. To calculate the ISS score, the score of the three most severely injured body regions is squared, and the results are added to produce the ISS score. The ISS was categorized as mild (1–15), moderate (16–24), or severe (>24).
Marshall MOD score
The Marshall MOD score (22) was calculated as an index of organ dysfunction. This score has six variables, including (a) the respiratory system (PO2/FIO2 ratio); (b) the renal system (serum creatinine concentration); (c) the hepatic system (serum bilirubin concentration); (d) the hematologic system (platelet count); (e) the central nervous system (Glasgow Coma Scale), and (f) the cardiovascular system—the pressure-adjusted HR. Marshall score was calculated according to Marshall et al. (22).
Analysis of inflammation biomarkers
Blood samples were collected into citrated tubes via central venous catheters within 24 h of admission and daily up to 7 days after injury. The blood samples were centrifuged, and plasma aliquots were stored in cryoprecipitate tubes at −80°C for subsequent analysis of inflammatory mediators. The human inflammatory MILLIPLEX MAP Human Cytokine/Chemokine Panel-Premixed 23 Plex (Millipore Corporation, Billerica, Mass) and Luminex 100 IS (Luminex, Austin, Tex) were used to measure plasma levels of IL-1β, IL-1 receptor antagonist (IL-1RA), IL-2, soluble IL-2 receptor α (sIL-2Rα), IL-4, IL-5, IL-6, IL-7, IL-8 (CCL8), IL-10, IL-13, IL-15, IL-17, interferon γ (IFN-γ), IFN-γ inducible protein (IP-10; CXCL10), monokine induced by IFN-γ (MIG; CXCL9), macrophage inflammatory protein 1α (MIP-1α; CCL3), MIP-1β, CCL4, MCP-1 (CCL2), granulocyte-macrophage colony-stimulating factor (GM-CSF), eotaxin (CCL11), and tumor necrosis factor α. The Luminex system was used in accordance to the manufacturer’s instructions. NO2−/NO3− was measured using the nitrate reductase/Griess assay (Cayman Chemical Co, Ann Arbor, Mich).
All data were analyzed using SigmaPlot 11 software (Systat Software, Inc, San Jose, Calif). Statistical difference between subcohorts was determined by either Student t test or χ2 test as appropriate. Group-time interaction of plasma inflammatory mediators’ levels between mild, moderate, and severe cohorts was determined by two-way analysis of variance (ANOVA). To quantify the differences among the statistically significant mediators, we calculated the area under the curve (AUC) using the mean values for each time point, then calculating moderate injury patients/mild injury patients’ and severe injury patients/mild injury patients’ AUC fold change. P < 0.05 was considered statistically significant for all analyses.
Dynamic Bayesian Network inference
Dynamic Bayesian Network (DyBN) inference was used to model the evolution of the probabilistic dependencies within a system over time. This analysis was carried out using MATLAB (The Math Works, Inc, Natick, Mass), using an algorithm adapted from Grzegorczyk and Husmeier (23) and revised recently by our group. In this analysis, inflammatory mediators were represented at multiple time points within the same network structure. In this approach, time was modeled discretely as in a discrete Markov chain. Each mediator was given a time index subscript indicating the time slice to which it belonged. Additional temporal dependencies were represented in a DyBN by edges between time slices. Each node in the network was associated with a conditional probability distribution of a variable that is conditioned upon its parents (upstream nodes). This particular network structure was used to assess the dominant inflammatory mediators and the probable interaction among various mediators, including possible feedback loops (18, 19, 23–25).
Demographics and clinical outcomes for the overall study cohort
The majority of 472 patients in the study population were males (70%), with a mean age of 48.4 ± 0.9 years and a mean ISS of 19.6 ± 0.5. These patients sustained blunt trauma in the form of MVC (57%), falls (21.6%), motorcycle crashes (11.2%), all-terrain vehicle accidents (ATVs; 2.5%), pedestrian run-over (1.3%), crush injury (0.9%), machinery injury (0.9%), cyclist injury (0.4%), and others (4.2%). The mean of ICU LOS was 6.9 ± 0.4 days, the mean hospital LOS was 12.7 ± 0.5 days, and the mean number of days on a mechanical ventilator was 2.9 ± 0.3 days.
Overall demographics and clinical outcomes of stringently matched subcohorts
To test our core hypothesis regarding differential trajectories and networks of systemic inflammation as a function of injury severity, we sought to derive stringently matched subcohorts that would be as similar as possible with regard to their basic demographics. As part of this process, we focused on the most common mechanisms of injury. Thus, from the previously described overall patient cohort, 147 patients were selected for this study: a subcohort of 49 patients with mild injury, a subcohort of 49 patients with moderate injury, and a subcohort of 49 patients with severe injury (see Materials and Methods). In this selection process, we focused on the MVC, falls, and motorcyclist injury groups, as those were the most common mechanisms of injury.
To determine whether the upper-range value of one group is not statistically significantly different from the lower-range value of the other group, we performed a statistical comparison on the mean ISS values of the mild versus moderate groups and the moderate versus severe groups (Table 1). As expected per our study design, ISS was significantly higher in the severe injury subcohort (P < 0.001) when compared with the moderate injury cohorts; similarly, ISS was significantly higher in the moderate injury group versus the mild injury group (P < 0.001). Overall, males were predominant in the mild, moderate, and severe injury subcohorts (67.3%) with no statistical difference in mean age (P = 0.8) among the groups. Supporting the notion that the derived subcohorts reflected the known outcomes associated with increasing injury severity (8, 26), the ICU LOS (P < 0.001), hospital LOS (P < 0.001), and days on mechanical ventilation (P < 0.001) were all statistically significantly longer in the severely injured cohort when compared with the mildly and moderately injured cohorts, respectively (Table 1). In addition, moderately injured patients exhibited longer ICU LOS, hospital LOS, and requirement for mechanical ventilation as compared with the mildly injured patients (Table 1).
Impact on distinct body regions as a function of injury severity
We next sought to compare the AIS in the stringently matched subcohorts to identify whether injuries in particular body regions could be associated with the worse outcomes posttrauma as a function of increasing ISS. Accordingly, we used the 2005 AIS version to describe injuries in specific anatomic regions (see Materials and Methods). Although the subcohorts were matched for age and were composed of similar sex ratios of patients injured by similar mechanisms of injury, statistically significant differences were observed in the head and neck (P < 0.001), chest (P < 0.001), abdomen (P = 0.001), and extremities (P = 0.002) in the severely injured subcohort when compared with mildly and moderately injured subcohorts (Fig. 1).
Prevalence of prehospital hypotension is higher as a function of injury severity
Several studies have reported a strong relationship between the severity of injury and the prevalence of prehospital hypotension, which in turn was associated with worse clinical outcomes (19). The prevalence of prehospital hypotension in our cohorts was 31% in the mildly injured subcohort, 41% in the moderately injured subcohort, and 61% in the severely injured subcohort (P < 0.001).
Elevated lactate and BD as a function of injury severity
Several biochemical parameters at admission in trauma patients have been shown to be associated with higher morbidity and mortality following injury. Accordingly, we evaluated HR, RR, T, lactate, BD, Hgb, Hct, platelet counts, and WBC count in our stringently matched subcohorts. This analysis showed that lactate (4.1 ± 0.3 vs. 2.8 ± 0.2 vs. 2.9 ± 0.3; P = 0.002; Fig. 2A) and BD (7.2 ± 0.6 vs. 4.3 ± 0.5 vs. 5.2 ± 0.7; P = 0.004; Fig. 2B) assessed upon admission were significantly different, respectively, in severely injured patients when compared with mildly or moderately injured patients. There were no significant differences among the three cohorts with regard to HR, RR, T, Hgb, Hct, platelet counts, or WBC counts upon admission (see Table, Supplemental Digital Content 1, at https://links.lww.com/SHK/A296).
Greater requirement for surgical interventions as a function of injury severity
Trauma patients either required surgical interventions to achieve hemodynamic stability, or they presented with simple or compound fractures that required either reduction and/or fixation. In the stringently matched subcohorts, we found that 14 (28.6%) of 49 mild injury patients, 20 (40.8%) of 49 moderate injury patients, and 28 (57%) of 49 severe injury patients underwent laparotomy to explore the source of bleeding and control it. Furthermore, 6 (12.2%) of 49 patients in the mild injury subcohort, 14 (28.6%) of 49 patients in the moderate injury subcohort, and 25 (51%) of 49 in the severe injury subcohort had either simple or compound fractures that required either fixation and/or reduction. Finally, 29 (59.2%) of 49 mild injury patients, 18 (36.7%) of 49 moderate injury patients, and 7 (14.3%) of 49 severe injury patients did not require any surgical intervention through their clinical course (P < 0.001). These data are depicted in Table 1.
Greater multiple organ dysfunction is a function of injury severity
The mild, moderate, and severe injury subcohorts differed in their degree of multiple organ dysfunction syndrome, as indicated by the Marshall MOD score (a well-validated index of dysfunction in multiple organ systems (27), which was calculated at each time point in which inflammation biomarkers were assessed. This analysis suggested that severely injured patients had a statistically significantly higher degree of organ dysfunction (P < 0.001) at day 1 and up to day 7 after injury when compared with the moderately and mildly injured patients, respectively (Fig. 3).
Prevalence of NI is higher as a function of injury severity
Susceptibility to NI depends on severity of injury, as well as the specific body region injured (17). Moreover, NI is associated with a longer hospital LOS and ICU LOS (17). In agreement with these prior studies, our data showed that the prevalence of NI increased significantly with severity of injury: 16% in the mildly injured subcohort, 29% in the moderately injured subcohort, and 45% in the severely injured subcohort (P = 0.008).
Different trajectories of systemic inflammation as a function of injury severity
Both blunt trauma itself and multiple organ dysfunction syndrome help drive systemic inflammation (28). Inflammation after injury is a complex, dynamic process that involves multiple body compartments, spills out into the systemic circulation, and defies simple characterization; we have utilized computational modeling to gain insights into this process (29,30). Accordingly, we next sought to define the dynamic patterns and networks of systemic inflammation as a function of injury severity. We first hypothesized that the dynamics of circulating inflammatory mediators would differ in the severe injury subcohort as compared with the mild and moderate injury subcohorts. We tested this hypothesis by obtaining an extensive time course of circulating inflammation biomarkers from close to the onset of injury and up to 7 days after injury in each of the injury severity subcohorts. Over the full 7 days’ time course, this analysis showed that circulating levels of IL-6 (P < 0.001), IL-7 (P = 0.04), IL-17 (P < 0.001), sIL-2Rα (P < 0.001), GM-CSF (P = 0.01), IP-10/CXCL10 (P = 0.03), MIG/CXCL9 (P = 0.03), and MCP-1/CCL2 (P = 0.02) were significantly higher in severely injured patients when compared with mildly and moderately injured patients (see Figure, Supplemental Digital Content 2, at https://links.lww.com/SHK/A297). An analysis of AUC, in which circulating inflammatory mediators were ranked according to the fold change (moderate injury patients/mild injury patients, and severe injury patients/mild injury patients), is shown in Table 2.
Next, we sought to focus on the dynamics of systemic inflammation over the first 24 h after injury in the stringently matched subcohorts, because most of the inflammatory changes happened in the early phase, influenced by different factors including magnitude of injury, resuscitation by fluids, blood components, and surgical interventions. The analysis of inflammation biomarkers for the first 24-h samples among the mildly, moderately, and severely injured subcohorts showed that circulating levels of IL-6 (P = 0.005), MCP-1/CCL2 (P < 0.001), IL-7 (P = 0.01), IL-17 (P = 0.018), IL-1RA (P = 0.031), and sIL-2Rα (P = 0.011) were significantly higher in severely injured patients when compared with mildly and moderately injured patients (see Figure, Supplemental Digital Content 2, at https://links.lww.com/SHK/A297). An analysis of AUC, in which circulating inflammatory mediators were ranked according to the fold change (moderate injury patients /mild injury patients, and severe injury patients/mild injury patients), is shown in Table 3.
Different dynamic networks of systemic inflammation as a function of injury severity
Based on these findings, we next hypothesized that the differences in the dynamic, systemic inflammatory response among the mild, moderate, and severe injury cohorts could be explained, at least in part, by differential expression of dynamic networks. We have recently demonstrated the utility of DyBN as a framework for biomarker discovery and to gain insight into acute inflammation in clinical settings (18,19,24). Based on DyBN, and similar to our prior studies in injured patients (18), we inferred that the chemokines MIG/CXCL9, IP-10/CXCL10, and MCP-1/CCL2 were central mediators of systemic inflammation in the mild injury subcohort, exhibiting self-feedback on their own production, with MIG/CXCL9 /CXCL9 being a central node of inflammation in these patients. This chemokine-based network appeared to affect the levels of IL-6, IL-8, (MIP-1α) (CCL3), and IL-1RA in mildly injured patients, whereas MCP-1/CCL2 affects the levels of IL-1RA, eotaxin (CCL11), and sIL-2Rα (Fig. 4A). In the moderate injury subcohort, DyBN inference suggested that IP-10/CXCL10, MIG/CXCL9, and MCP-1/CCL2 exhibited self-feedback on their own production, with MCP-1/CCL2 being a central node in the dynamic circulating inflammatory response of these patients. This MCP-1/CCL2 subnetwork was inferred to affect IL-6, IL-1RA, sIL-2Rα, IL-10, MIP-1α (CCL3), and eotaxin (CCL11). This analysis also suggested that MIG/CXCL9 affects the levels of IFN-α and IL-6, whereas IP-10/CXCL10 affects the levels of IL-10, MIP-1α (CCL3), and eotaxin (CCL11) (Fig. 4B). In the severe injury subcohort, DyBN suggested that, as in the other cohorts, IP-10/CXCL10, MIG/CXCL9, and MCP-1/CCL2 exhibited self-feedback on their own production. In these patients, MCP-1/CCL2 was inferred to affect IL-6, eotaxin (CCL11), NO2−/NO3−, MIP-1β (CCL3), and sIL-2Rα; MIG/CXCL9 was inferred to affect IL-6 and eotaxin (CCL11); and IP-10/CXCL10 was inferred to affect IL-6 and sIL-2Rα (Fig. 4C).
Taken together, this analysis of dynamic networks of systemic inflammation suggested that, among multiple mediators, the key injury-associated cytokine IL-6 (31) is affected by three chemokines (MCP-1/CCL2, MIG/CXCL9, and IP-10/CXCL10) in severely injured patients. In contrast, IL-6 appeared to be affected by only two chemokines (MCP-1/CCL2 and MIG/CXCL9) in moderately injured patients, while being regulated solely by MIG/CXCL9 in mildly injured patients.
Multiple factors, including age (32), sex (33), and the presence of hypotension (19) can affect the outcomes of blunt trauma. However, the most common factor to consider when deciding how to best target care is the severity of the injury itself. This core consideration has driven the development and refinement of multiple trauma scoring systems over the last few decades, among which ISS remains the most commonly used (20) for anatomical injury assessment. Although imperfect in certain circumstances, and worse than other scoring systems such as the Trauma Injury Severity Score with regard to outcome prediction, ISS correlates linearly with outcomes in the posttraumatic clinical course as well as with the late complications. Despite its limitations and compared with other injury severity scores, the ISS comes closest to quantifying the degree of externally driven injury independently of the host’s physiological responses to that injury. Furthermore, ISS affects the final destination for patients after hospitalization, as there is a relation between the severity of the injury (calculated via the ISS) and the functional status of the patients following hospitalization (34).
Another key factor that drives outcomes after injury is the systemic inflammatory and pathophysiologic response that ensues. Trauma and hemorrhage trigger a complex cascade of events associated with alterations in inflammatory/immune responses that are largely orchestrated by cytokines, chemokines, damage-associated molecular patterns molecules, and free radicals (35, 36). The dynamic interplay among these factors, which occurs at multiple scales from the molecular to the whole organism and is associated with a massive reprogramming of gene expression (15), is a complex cascade of events. Thus, focusing on single mediators or pathways as a “magic bullet” approach is likely to be nonproductive; instead, a systems approach is needed to address this complexity (37).
We and others have used both data-driven and mechanistic computational modeling approaches to address this complexity and to gain both basic and translational insights in trauma, hemorrhage, and related phenomena such as sepsis (29, 30). Our focus in this study was to address a basic question regarding whether the degree or character of systemic inflammation changes as a function of injury severity, because there is a lack of consensus in the literature with regard to the impact on inflammation of increasing injury severity. Our hypothesis in the present study was the differential immune/inflammatory mechanisms are set in motion by injury of varying severity, in a manner associated with differential clinical outcomes. The systemic circulation is both a locus for this dysregulated inflammation and a convenient compartment for assessing the inflammatory response (17–19). Accordingly, we sought to define the dynamic trajectories and networks of inflammation as a function of injury severity and to correlate these differences with key in-hospital and discharge-related outcomes.
To carry out this study, we first obtained highly granular clinical and inflammation biomarker data, with a focus on extensive early sampling (within 24 h and out to 7 days after injury). Second, we generated stringently, post hoc–matched cohorts of mildly, moderately, and severely injured patients, to reduce the impact of confounding factors while still retaining the core clinical characteristics. These subcohorts have face validity precisely because they reproduce known findings with regard to the generally worse clinical outcome as a function of injury severity (derived from multiple studies that did not filter out the confounding variables we removed, such as alcohol intoxication). These two goals were integral to the design of our clinical protocol, facilitating a data-driven, precision-medicine approach to addressing our core questions, “Does systemic inflammation increase as a function of injury severity, or does the character of this dynamic inflammatory response differ significantly as injury becomes greater?”
To address this question, we first confirmed key prior observations about the impact of different injury severity in blunt trauma patients, namely, longer ICU and total LOS, greater requirement for mechanical ventilation, higher prevalence of prehospital hypotension, a higher degree of organ dysfunction, and a higher prevalence of NI, which were statistically different between the subcohorts. Furthermore, we examined several physiological parameters independent of ISS and were unable to identify any significant differences among the three stringently matched groups in T, HR, RR, blood pressure, and Glasgow Coma Scale score. These outcomes, which in some cases varied linearly as a function of injury severity, were correlated with differential inflammatory trajectories and dynamic networks.
Our data suggest that the intensity of activation of multiple inflammatory pathways, including those pathways driven by macrophages/neutrophils as well as lymphoid cells, roughly increases as a function of injury severity. This result is in agreement with prior studies (35). In many prior studies of trauma/hemorrhage in both animals and patients, the cytokine IL-6 has persisted as a key inflammatory mediator and biomarker (38). Prior studies from our groups have leveraged various computational modeling tools to discern the importance of chemokines in driving the inflammatory response—and IL-6 in particular—in both mice and blunt trauma patients (39). In this context, this study reinforces a key role for the cytokine IL-6 (as gleaned from both animal and clinical studies), but does so in a manner that puts this cytokine in the dynamic context of other inflammatory mediators, that is, finding that differential network connectivity is a key feature of increasing injury severity. Most importantly, and by identifying key mediators in each group, our data-driven approach could increase our understanding of the acute, systemic inflammatory response to traumatic injury and thereby aid in the implementation of innovative therapies that target specific components of the mediator cascade.
In the present study, we carried out DyBN, a network discovery method that allows for inference of feedback in network nodes. We have used this method previously as a novel biomarker framework in pediatric acute liver failure (24) and to suggest key dynamic networks of inflammation in blunt trauma and spinal cord injury (18,19). Our analysis is perhaps best explained by stating that based on differential chemokine interactions over time there is a resultant impact on the dynamic production of IL-6, which in turn might impact the peak systemic levels of IL-6. These results support our prior studies combining in vitro, in silico, and clinical studies, in which MCP-1/CCL2 was suggested to regulate IL-6 production (39). Further computational modeling studies from our group suggest that this injury-graded network can account for some aspects of the overall systemic inflammatory response as well as patient outcomes (Azhar et al., in preparation). Nonetheless, further mechanistic studies are needed in order to prove the hypothesis regarding chemokine-based regulation of systemic IL-6.
We recognize that there are several limitations in our study. First, this study was performed at a single, level I trauma center and thus may not be generalizable or pertinent to other centers with differing admission demographics, injury characteristics, or management practices. This issue warrants additional, similar studies in other trauma centers to validate the results suggested from the current study. The use of the ISS as a surrogate for the extent of injury may itself be considered as a limitation, as many other scoring systems are available for quantifying the overall physiological burden associated with trauma/hemorrhage. Indeed, we have demonstrated that additional components beyond the injury itself (eg, spinal cord injury , NI , or hypotension ) drive complex changes in dynamic networks of systemic inflammation in the context of blunt trauma.
Although our three study subcohorts were matched stringently with regard to age and mechanism of injury, as well as comprising a similar mix of males and females, statistically significant differences were observed in multiple body regions in the severely injured subcohort when compared with mildly and moderately injured groups. Thus, we cannot fully rule out that it is these differences in injured body regions, rather than the ISS as a composite score thereof, which drove the differential inflammatory responses and clinical outcomes observed. Moreover, significant differences were observed in the prevalence of prehospital hypotension and NI among cohorts, which may play a role in the different patterns of the inflammatory response, which cannot be ignored. Another important limitation of this retrospective study is the potential impact of blood transfusion and surgical interventions on the temporal dynamics of the inflammatory response. We note that these interventions are by necessity an intrinsic element of clinical care for management of trauma patients with evidence of blood loss. We suggest that these patients who received this type of interventions reflect patients with persistent hemodynamic instability and that by excluding these patients we will be missing the opportunity to capture the early dynamic changes that accompany persistent hypotension. Another limitation concerns the set of inflammatory mediators assessed in this study, which represent a broad cross section of innate and lymphoid mediators and thus interrogate most pathways examined to date in blunt trauma. Furthermore, the mediators represented in this beadset interrogate most pathways examined to date in blunt trauma, although additional mediators need to be studied in order to examine their potential role in following injury. Finally, we note that data-driven modeling relies on available data and as such depends on the quality of those data. These tools do not provide any direct mechanistic information about the biology beneath it; however, they are suggesting possible interactions among inflammatory mediators.
In conclusion, the current study demonstrates the presence of differential, injury-graded early systemic inflammatory responses, which may explain why IL-6 may be a persistently useful inflammation biomarker. These inflammatory responses are associated fairly tightly with significantly different clinical outcomes. We suggest that by using extensive early sampling coupled to stringent post hoc matching of subcohorts, we were able to identify key networks that characterize the trauma-induced inflammatory response and that we identified differential dynamic network connectivity as a key output of increasing injury severity. Using this methodology, we identified key networks that characterize the trauma-induced inflammatory response in a manner that was as free from confounding comorbidities and other clinical factors that could obscure these findings. By identifying key mediators in each group, our data-driven approach could facilitate the implementation of innovative therapies that target specific components of the mediator cascade in an attempt to improve patients’ outcomes following trauma. In addition, and in the context of precision medicine, we suggest that analysis of the dynamic networks (by DyBN) during the initial 24 h after injury could eventually allow for early stratification of patients who develop critical illness and hence improve allocation of resource intensive care.
The authors thank Nabil Azhar (Department of Department of Computational and Systems Biology, University of Pittsburgh) for developing a MATLAB-integrated interface, which was used to run the DyBN analysis.
1. Kauvar DS, Lefering R, Wade CE: Impact of hemorrhage on trauma outcome: an overview of epidemiology, clinical presentations, and therapeutic considerations. J Trauma
60 (Suppl 6): S3–S11, 2006.
2. Wardle TD: Co-morbid factors in trauma patients. Br Med Bull
55 (4): 744–756, 1999.
3. Aldemir M, Tacyildiz I, Girgin S: Predicting factors for mortality in the penetrating abdominal trauma. Acta Chir Belg
104 (4): 429–434, 2004.
4. Pape HC, Tsukamoto T, Kobbe P, Tarkin I, Katsoulis S, Peitzman A: Assessment of the clinical course with inflammatory parameters. Injury
38 (12): 1358–1364, 2007.
5. Lecky F, Woodford M, Edwards A, Bouamra O, Coats T: Trauma scoring systems and databases. Br J Anaesth
113 (2): 286–294, 2014.
6. Russell RJ, Hodgetts TJ, McLeod J, Starkey K, Mahoney P, Harrison K, Bell E: The role of trauma scoring in developing trauma clinical governance in the Defence Medical Services. Philos Trans R Soc Lond B Biol Sci
366 (1562): 171–191, 2011.
7. Vucovic D, Lazarevic D, Miskovic G, Stefanovic B: A scoring system for polytrauma patients [in Croatian]. Acta Chir Iugosl
46 (1–2): 17–30, 1999.
8. Baker SP, O’Neill B, Haddon W Jr, Long WB: The Injury Severity Score: a method for describing patients with multiple injuries and evaluating emergency care. J Trauma
14 (3): 187–196, 1974.
9. Champion HR, Sacco WJ, Copes WS, Gann DS, Gennarelli TA, Flanagan ME: A revision of the Trauma Score. J Trauma
29 (5): 623–629, 1989.
10. Boyd CR, Tolson MA, Copes WS: Evaluating trauma care: the TRISS method. Trauma Score and the Injury Severity Score. J Trauma
27 (4): 370–378, 1987.
11. Palmer C: Major trauma and the Injury Severity Score—where should we set the bar? Annu Proc Assoc Adv Automot Med
51: 13–29, 2007.
12. Rutledge R, Fakhry S, Rutherford E, Muakkassa F, Meyer A: Comparison of APACHE II, Trauma Score, and Injury Severity Score as predictors of outcome in critically injured trauma patients. Am J Surg
166 (3): 244–247, 1993.
13. Hurr H, Hawley HB, Czachor JS, Markert RJ, McCarthy MC: APACHE II and ISS scores as predictors of nosocomial infections in trauma patients. Am J Infect Control
27 (2): 79–83, 1999.
14. Ertel W, Keel M, Bonaccio M, Steckholzer U, Gallati H, Kenney JS, Trentz O: Release of anti-inflammatory mediators after mechanical trauma correlates with severity of injury and clinical outcome. J Trauma
39 (5): 879–885, 1995; discussion 885–887.
15. Xiao W, Mindrinos MN, Seok J, Cuschieri J, Cuenca AG, Gao H, Hayden DL, Hennessy L, Moore EE, Minei JP, et al.: Inflammation and Host Response to Injury Large-Scale Collaborative Research Project: a genomic storm in critically injured humans. J Exp Med
208 (13): 2581–2590, 2011.
16. Vodovotz Y, Billiar TR: In silico modeling: methods and applications to trauma and sepsis. Crit Care Med
41 (8): 2008–2014, 2013.
17. Namas RA, Vodovotz Y, Almahmoud K, Abdul-Malak O, Zaaqoq A, Namas R, Mi Q, Barclay D, Zuckerbraun B, Peitzman AB, et al.: Temporal patterns of circulating inflammation biomarker networks differentiate susceptibility to nosocomial infection following blunt trauma in humans [published online ahead of print November 3, 2014]. Ann Surg
18. Zaaqoq AM, Namas R, Almahmoud K, Azhar N, Mi Q, Zamora R, Brienza DM, Billiar TR, Vodovotz Y: Inducible protein-10, a potential driver of neurally controlled interleukin-10 and morbidity in human blunt trauma. Crit Care Med
42 (6): 1487–1497, 2014.
19. Almahmoud K, Namas RA, Zaaqoq AM, Abdul-Malak O, Namas R, Zamora R, Sperry J, Billiar TR, Vodovotz Y: Prehospital hypotension is associated with altered inflammation dynamics and worse outcomes following blunt trauma in humans. Crit Care Med
43 (7): 1395–1404, 2015.
20. Glance LG, Osler TM, Mukamel DB, Meredith W, Dick AW: Expert consensus vs empirical estimation of injury severity: effect on quality measurement in trauma. Arch Surg
144 (4): 326–332; discussion 332, 2009.
21. Gennarelli TA, Wodzin E: AIS 2005: a contemporary injury scale. Injury
37 (12): 1083–1091, 2006.
22. Marshall JC, Cook DJ, Christou NV, Bernard GR, Sprung CL, Sibbald WJ: Multiple organ dysfunction score: a reliable descriptor of a complex clinical outcome. Crit Care Med
23 (10): 1638–1652, 1995.
23. Grzegorczyk M, Husmeier D: Improvements in the reconstruction of time-varying gene regulatory networks: dynamic programming and regularization by information sharing among genes. Bioinformatics
27 (5): 693–699, 2011.
24. Azhar N, Ziraldo C, Barclay D, Rudnick DA, Squires RH, Vodovotz Y, and for the Pediatric Acute Liver Failure Study Group: Analysis of serum inflammatory mediators identifies unique dynamic networks associated with death and spontaneous survival in pediatric acute liver failure. PLoS One
8 (11): e78202, 2013.
25. Langmead CJ: Generalized Queries and Bayesian Statistical Model Checking in Dynamic Bayesian Networks: Application to Personalized Medicine
. Pittsburgh, PA: School of Computer Science, Carnegie Mellon University, 2009.
26. Salottolo K, Settell A, Uribe P, Akin S, Slone DS, O’Neal E, Mains C, Bar-Or D: The impact of the AIS 2005 revision on injury severity scores and clinical outcome measures. Injury
40 (9): 999–1003, 2009.
27. Sauaia A, Moore EE, Johnson JL, Ciesla DJ, Biffl WL, Banerjee A: Validation of postinjury multiple organ failure scores. Shock
31 (5): 438–447, 2009.
28. Jastrow KM 3rd, Gonzalez EA, McGuire MF, Suliburk JW, Kozar RA, Iyengar S, Motschall DA, McKinley BA, Moore FA, Mercer DW: Early cytokine production risk stratifies trauma patients for multiple organ failure. J Am Coll Surg
209 (3): 320–331, 2009.
29. An G, Nieman G, Vodovotz Y: Computational and systems biology in trauma and sepsis: current state and future perspectives. Int J Burns Trauma
2 (1): 1–10, 2012.
30. Vodovotz Y, Billiar TR: In silico
modeling: methods and applications to trauma and sepsis. CritCare Med
41: 2008–2014, 2013.
31. Jawa RS, Anillo S, Huntoon K, Baumann H, Kulaylat M: Interleukin-6 in surgery, trauma, and critical care part II: clinical implications. J Intensive Care Med
26 (2): 73–87, 2011.
32. Adams SD, Cotton BA, McGuire MF, Dipasupil E, Podbielski JM, Zaharia A, Ware DN, Gill BS, Albarado R, Kozar RA, et al.: Unique pattern of complications in elderly trauma patients at a level I trauma center. J Trauma Acute Care Surg
72 (1): 112–118, 2012.
33. Sperry JL, Friese RS, Frankel HL, West MA, Cuschieri J, Moore EE, Harbrecht BG, Peitzman AB, Billiar TR, Maier RV, et al.: Male gender is associated with excessive IL-6 expression following severe injury. J Trauma
64 (3): 572–578; discussion 578–579, 2008.
34. Weninger P, Aldrian S, Koenig F, Vecsei V, Nau T: Functional recovery at a minimum of 2 years after multiple injury—development of an outcome score. J Trauma
65 (4): 799–808; discussion 808, 2008.
35. Lenz A, Franklin GA, Cheadle WG: Systemic inflammation after trauma. Injury
38 (12): 1336–1345, 2007.
36. Osuka A, Ogura H, Ueyama M, Shimazu T, Lederer JA: Immune response to traumatic injury: harmony and discordance of immune system homeostasis. Acute Med Surg
1 (2): 63–69, 2014.
37. Buchman TG, Cobb JP, Lapedes AS, Kepler TB: Complex systems analysis: a tool for shock research. Shock
16 (4): 248–251, 2001.
38. Gebhard F, Pfetsch H, Steinbach G, Strecker W, Kinzl L, Bruckner UB: Is interleukin 6 an early marker of injury severity following major trauma in humans? Arch Surg
135 (3): 291–295, 2000.
39. Ziraldo C, Vodovotz Y, Namas RA, Almahmoud K, Tapias V, Mi Q, Barclay D, Jefferson BS, Chen G, Billiar TR, et al.: Central role for MCP-1/CCL2 in injury-induced inflammation revealed by in vitro
, in silico
, and clinical studies. PLoS One
8 (12): e79804, 2013.