Traumatic brain injury with concomitant injury to the spleen: characteristics and mortality of a high-risk trauma cohort from the TraumaRegister DGU®

Traumatic brain injury (TBI) remains a major cause of mortality and disability particularly in young adults [1]. Local and systemic inflammation has been identified as an important contributor to secondary brain injury, which leads to additional downstream damage and represents a potential therapeutic target for neuroprotective approaches [2]. Peripheral immune mechanisms lead to systemic release of proinflammatory mediators and entry of immune cells into the CNS, particularly in the context of blood–brain barrier disruption [3, 4]. Peripheral immune organs like the spleen, which serves as a storage site for immune cells like monocytes [5], may be major contributors to peripheral immune responses. Indeed, a proinflammatory contribution of the spleen has already been demonstrated in the context of cerebral ischemia [6‐8]. This important role of the spleen related to systemic immunoresponses has been also extrapolated to experimental TBI [9]. In rodent models of TBI, splenectomy led not only to a suppression of proinflammatory cascades including reduction in inflammatory cytokines like TNF-α and IL-6 but also was associated with better survival along with improved cognitive functions [10, 11]. Of note, experimental data suggest that the adaptive immunity does not seem to considerably contribute to this immune pathomechanism [12]. However, analysis of spleen cells showed an increase in splenic lymphocytes including T cells within 24 h after experimental TBI, which was reversed in the context of CD74 deficiency leading to a smaller lesion size and decreased neurodegeneration [13].

Despite the interesting nature of this preclinical knowledge and the potential therapeutic implications, translational exploration of the effect of splenectomy in patients in a controlled prospective approach appears not feasible. Structured trauma registries may hold the potential to narrow this gap in knowledge by providing relevant data on surgical treatments and outcomes across a range of traumatic impacts. Based on the preclinical data and consequent hypothesis that splenectomy improves survival after TBI, Teixeira and co-workers have queried the National Trauma Data Bank (NTDB) to assess the clinical role of splenectomy in moderate-to-severe TBI patients with splenic injury and exploratory laparotomy [14]. In their study, however, the authors observed an increased mortality with splenectomy, favoring organ preservation over resection in TBI patients with concomitant injury to the spleen. But the validity of this conclusion has been questioned [15]. Potentially confounding factors, such as hemodynamic data, are not routinely captured into the NTDB and could, thus, not be included into the analysis.

This study, therefore, aims to provide a more in-depth description of a trauma cohort suffering both from moderate-to-severe TBI with concomitant spleen injury based on the German TraumaRegister DGU^® (TR-DGU). The TR-DGU prospectively collects patient-level data on demographics, injury pattern, comorbidities, pre- and in-hospital management including surgical interventions, e.g. splenectomy, as well as outcomes in four consecutive intervals from the injury until hospital discharge. The assessment of these data, at least retrospectively, allows for systematic analysis of the course of TBI patients with or without splenectomy. Given the previous discrepancy between preclinical and clinical data, the present study is driven by the hypothesis that complete removal of splenic tissue reduces mortality in patients with moderate-to-severe TBI. Accordingly, we provide an observational cohort study with insights on demographics, clinical characteristics including hemodynamic parameters and in-hospital mortality of a selected high-risk cohort of trauma patients with TBI and injury to the spleen. Based on these data, the effect of splenectomy on mortality is tested after adjustment for confounding factors.

The study cohort consisted of 1114 patients who met the set inclusion criteria (Fig. 1). Out of these, 328 (29.4%) had undergone early splenectomy. Among the remaining 786 patients were 168 patients (15.1%) who received either a spleen preserving intervention, or a late splenectomy (n = 13). Table 1 displays the rate of splenectomy against the magnitude of injury to the spleen reflected by the AIS_spleen. The majority of splenectomies (95.9%) were performed within two days after initial trauma (Table 2).

The groups shared a male-dominant sex distribution (about 70%) and a similar mean age (about 40 years). The most common trauma mechanism was traffic related (75.2%) and the most common concomitant injury was thoracic trauma. The percentage of patients with an initial GCS ≤ 8 was higher in patients who were splenectomized (49.4% vs. 57.3%, p = 0.017). Patients with early splenectomy had a much higher Injury Severity Score (median: 34 vs. 44, p < 0.001) and a worse RISC-II prognosis (32.7% vs. 20.8%, p < 0.001). These patients were also more frequently hypotensive upon admission (19.8% vs. 38.0%, p < 0.001) and in need for blood transfusion (30.3% vs. 61.0%, p < 0.001).

The mortality was 20.7% in the splenectomy group and 10.3% in the control group without early splenectomy. Crude neurological outcome at discharge was documented as favorable in 65.1% of splenectomized survivors and in 71.1% of control cases (p = 0.083). The median length of stay (LOS) in hospital was similar in both groups (23 vs. 25 days, p = 0.25) but median ICU stay (14 vs. 16 days, p = 0.014) and duration of mechanical ventilation (7 vs. 14 days, p < 0.001) were longer in the splenectomized group. Table 3 provides an overview about the demographic, clinical and outcome data of the two subgroups.

After adjustment for the RISC-II score, injury severity of the spleen, and other surgical interventions in a logistic regression model (Table 4), splenectomy was not found to exert an effect on in-hospital mortality (OR 1.29 (0.67–2.50), p = 0.45). RISC-II score was the only covariable demonstrating a significant effect in this model.

The sub-cohort of patients with surgical therapy related to the spleen was further described with a propensity score analysis. Within this sub-cohort, 107 patients with and without splenectomy, respectively, could be matched according to the propensity score. No statistical differences in mortality or secondary outcome parameters between the two groups were observed (Table 5).

The spleen is a major lymphatic organ serving as a storage site for lymphocytes as well as monocytes, which can rapidly be deployed to tissue injuries [5, 13]. A protective effect of splenectomy against ischemic damage and associated proinflammatory reactions in different organs has already been described [19, 20]. With regards to cerebral injuries, an inflammatory involvement of the spleen in the pathophysiology of ischemic brain lesions has been demonstrated [8, 21]. Activated spleen cells have been shown to secrete enhanced levels of inflammatory mediators within the first 22 h after experimental stroke [22]. However, it has also been demonstrated that this is followed by splenic atrophy with a concomitant state of immunosuppression four days after stroke induction [23]. Splenectomy prior or splenic irradiation after experimental stroke induction reduced infarction volumes and cerebral immune cell counts in rats [6, 7]. Investigation of the role of splenectomy in experimental TBI showed similar findings of an ameliorated peripheral inflammatory reaction [9‐11]. Immediate posttraumatic splenectomy led to an improved survival and cognitive outcome in rats [10]. The neuroprotective effect of splenectomy has been associated with a suppression of the proinflammatory mitogen-activated protein kinase signaling pathway in experimental TBI [11]. In summary, these preclinical findings showed promising potential of splenectomy to improve secondary brain injury by suppression of inflammation after acute cerebral injury. Moreover, splenectomy may also imply an earlier and more definitive bleeding control preventing a scenario of hemorrhagic shock due to occult hemorrhage, which is detrimental in the setting of polytrauma [24].

Besides the proinflammatory role of the spleen after TBI, a cholinergic anti-inflammatory pathway leading to spleen-dependent immunosuppression and nosocomial infections, particularly pneumonia, after TBI has also been described [25]. However, a recent clinical study contradicted this mechanism in humans by showing that splenectomy did not mitigate the risk of pneumonia in TBI patients [26]. This also underlines the importance of translationally confirming preclinical with clinical findings to further elucidate the actual human pathophysiology.

In the present study, we describe a trauma cohort with moderate-to-severe TBI and concomitant splenic injury. The total rate of splenectomy was 29.4% and markedly increased with higher magnitude of injury as reflected by the AIS_spleen. Congruent with this high rate of splenectomy in AIS_spleen 4&5 patients, TBI has previously been identified as a predictor for failure of nonoperative management in high grade splenic injury [27]. Splenectomized patients were about twice as likely to receive RBC transfusions as compared to the overall study cohort. Besides indicating a hemodynamically more unstable group, the role of RBC transfusion in the setting of TBI itself remains ambiguous with particularly liberal transfusion management being associated with worse outcomes [28, 29]. In general, patients who had received an early splenectomy were more severely injured and showed worse crude outcomes. This observation corresponds to the work by Teixeira et al. [14]. In their work based upon NTDB data, splenectomy remained associated with higher mortality also after adjustment for variables including age, hypotension at admission, GCS, AIS_abdomen, ISS, solid-organ or hollow-viscus injury and grade of splenic injury. Based on our logistic regression model and study cohort, we were not able to show a beneficial effect of splenectomy on outcome either. But in contrast to the NTDB analysis, we did not observe a negative effect on in-hospital mortality associated with splenectomy either. A possible reason might be different inclusion criteria given that only patients who had received an exploratory laparotomy were included in the study by Teixeira et al. [14]. However, we did not detect a significant difference in outcome in a propensity score subgroup analysis between patients with splenectomy or other splenic surgery either. Moreover, the difference in the effect of splenectomy might also be attributed to usage of the RISC-II score for adjustment. The RISC-II score allows a comprehensive characterization of the individual patient at the time of admission [18]. Similar to the regression model of Teixeira et al., it also considers the variables: age, blood pressure and overall trauma severity scales (AIS severity level of worst, second-worst and head injury). In addition to the motor function of GCS, pupil status is included as a further TBI-specific clinical sign. Additionally, besides pre-injury health status, important circulation specific parameters such as acidosis, coagulation, hemoglobin and CPR are included as well. A later deterioration, however, is not covered by the RISC-II and thus, splenectomy might be an indicator for such a dynamic. Thus, the conflicting findings from two studies can be attributed to insufficient patient adjustment by variables not considered or available in the NTDB dataset particularly displaying hemodynamic instability, which represents a major aspect in patients with severe splenic injury [15]. In this context, hypotension has been demonstrated to be predictive of mortality and functional outcome in TBI, and adherence of relatively high thresholds was associated with favorable outcomes [30]. Even though mortality was adjusted for hypotension at admission by Teixeira et al., possible episodes of hypotension particularly in the field and during transportation may not have been displayed; whereas, this would have been represented by variables of the RISC-II score like acidosis, coagulation or hemoglobin [18].

A limitation of the present study and a general issue of large-scale databases is the risk of incomplete or incorrect data despite various electronic plausibility checks when entering online data, which may have affected the presented results. Evaluation and documentation of source data are subject to interrater variability influenced by training background and local conventions. Patients documented only via the basic data set of the TR-DGU were not considered because no information about surgical procedures is collected for these cases. Possibly, this leads to a bias of higher representation of high-level trauma center in this study since the basic data set is mostly used by smaller hospitals. Nevertheless, the evaluated dataset is still among the largest sets available and provided a sufficient patient number for statistical analyses. The presented findings are only generalizable on cohorts within a comparable health care system and with similar inclusion and exclusion criteria. Notably, as in the NTDB study [14], all patients who died within the first 48 h have been excluded to focus on the effect of splenectomy on secondary brain injury and to control for the self-fulfilling prophecy effect of initial withdrawal of life-sustaining therapy due to high severity of primary trauma. Since inflammatory reaction exhibits temporal changes of cytokine and chemokine concentrations implicating also functional alterations [22], a potential confounding factor is the timing of splenectomy as also addressed by Teixeira et al. [14]. We have, therefore, focused on the role of early splenectomies only. Moreover, this study only looked at discharge mortality and neurologic outcome, and it remains unknown how long-term outcomes will develop at later timepoints. A major limitation is also that early neurologic outcome is based on the Glasgow Outcome Scale, which represents a rather crude assessment tool that was designed to determine long-term outcomes [16]. This scale might not be sensitive enough to detect subtle differences in the early neurological and functional status of a patient and therefore does not allow to investigate the effect of splenectomy on neurological outcome with scientific certainty. Access to scales of more complexity like the Rancho Los Amigos Scale would be advantageous for future studies [31].

The present main results and conclusions of this study are based on a logistic regression model with clear limitations. It is understood that such data are generally to be interpreted with caution and should not uncritically guide towards a change of paradigms without further investigation. Alike, this study cannot fully reject the concept of a detrimental peripheral inflammatory reaction which has previously been shown by experimental data. It seems natural that a direct comparison between a polytraumatized patient and a splenectomy under highly controlled experimental conditions without actual trauma other than TBI is challenging. It needs to be considered, that in the setting of polytrauma, also other injury factors like tissue damage and necrosis [32, 33] or hemorrhagic shock and subsequent massive blood product substitution [34] may drive peripheral inflammation and lead to a systemic inflammatory response syndrome. Despite extensive adjustment for confounding factors, the effect calculated for splenectomy based on the real-life data from a registry presumably still covers both the impact of the intervention itself—in which we are interested and on which the experimental data are based on—and also the critical overall situation which led to the splenectomy. A discrepancy between registry and preclinical studies might, therefore, be due to unknown variables we cannot adjust for as well as to a temporal bias given that we adjusted with the RISC-II score only for data collected at admission. However, a potential beneficial effect on in-hospital mortality seemed not strong enough to be measurable by our methodology. A prospective interventional trial, which would assess this concept in a more controlled fashion, appears not feasible for obvious ethical reasons.

In a cohort of patients with moderate-to-severe TBI and concomitant splenic injury, patients undergoing splenectomy demonstrate a more severe injury pattern, more compromised hemodynamic status and higher in-hospital mortality than patients without splenectomy. Adjustment for confounding factors reveals that the splenectomy procedure itself is not independently associated with survival. Matched pairs of patients with either splenic preservation surgery or splenectomy demonstrate no differences in mortality.