Intracranial lesion features in moderate-to-severe traumatic brain injury: relation to neurointensive care variables and clinical outcome

Traumatic brain injury (TBI) leads to a significant burden of mortality and morbidity worldwide [4, 30]. The post-traumatic intracranial hemorrhagic lesion type, size, mass effect, and evolution significantly affect the clinical course and outcome for the individual patient [13, 17, 18, 29]. Neurosurgical care in TBI is devoted to detection and emergency evacuation of significant intracranial hemorrhages to avoid brain herniation and neurointensive care (NIC) to improve the cerebral environment by managing systemic and cerebral physiology [11, 22]. Specifically, elevated intracranial pressure (ICP), low cerebral perfusion pressure (CPP), and disturbances in cerebral pressure autoregulation (PRx) are important predictors of poor outcome and are also potentially treatable insults during NIC [6, 10, 27]. There is a general understanding that large traumatic intracranial hemorrhages contribute to a more complicated clinical course and contribute to worse derangements in cerebral physiology (ICP, CPP, and PRx), both preoperatively due to the volume effect of the lesion as well as postoperatively, e.g., related to secondary brain swelling. However, only a few studies [34, 35] have in detail investigated the association of specific intracranial hemorrhagic types (epidural hematoma (EDH), acute subdural hematoma (ASDH), subarachnoid hemorrhage (SAH), intraventricular hemorrhage (IVH), and contusions) and their sizes, on the monitoring and treatment variables of the acute phase in the NIC. This is of interest in order to better understand how in-depth analyses of early radiological imaging data may be used to understand the immediate clinical course following moderate to severe TBI, which is often challenging to predict. In addition, attentive NIC strives to keep physiological variables within optimal ranges to avoid secondary brain insults [11, 22] and it remains to be determined to what extent early intracranial hemorrhage features affect the success rate of keeping the patients within these NIC targets. Thus, by means of granular volumetric intracranial hemorrhage assessments, detailed information of the clinical course, and high-frequency physiological data from the NIC in a large cohort of TBI patients, we aimed to explore the role of specific intracranial hemorrhage features (type of lesion, volumes, mass effect, and evolution) on the clinical course, derangements in cerebral physiology (ICP, CPP, and PRx) during NIC, and long-term clinical outcome. We hypothesized that large lesions of particularly EDHs, ASDHs, and contusions would require emergency evacuation, but would not lead to more ICP and CPP insults since these variables were actively optimized during vigilant NIC. PRx was not actively treated and was hypothesized to be more disturbed in association to larger hemorrhage volumes and specifically related to SAH and ASDHs [1, 34].

In this study, including 385 patients with moderate to severe TBI, we found that tSAH and hemorrhagic progression predicted more ICP problems during the first week post-injury. Hemorrhagic progression and the need for emergency neurosurgery were also associated with more CPP insults, whereas severe mass effect with compressed/obliterated basal cisterns independently predicted a more disturbed cerebral pressure autoregulation. Obliterated basal cisterns, presence of tSAH, and larger ASDHs and contusions were independently associated with worse clinical outcome. Although these results were to some extent expected, our study provides a granular analysis on the effect of specific intracranial hemorrhage lesion type, mass effect, and evolution in the early course on clinical trajectories and neurological recovery. Future efforts should aim to integrate such data into more sophisticated machine learning models to aid the clinician to better anticipate emerging secondary insults and to more accurately predict clinical outcome in TBI.

The patients with large EDHs were operated with craniotomy to a larger extent, reflecting the natural course of EDHs as rapidly evolving lesions that require emergency evacuation to avoid brain herniation [5]. However, they did not require last-tier ICP treatments (thiopental and DC) as often and were not more prone to develop ICP insults. Since the ICP monitor was generally inserted at the time of hematoma evacuation, these results primarily support that postoperative ICP problems were not very frequent, possibly thanks to timely surgery before secondary brain injury had occurred. Presence of/larger EDHs was associated with favorable outcome in univariate analyses but not in the multiple regression analysis. Since an EDH is extra-cerebral and not primarily a cerebral injury, it is a more treatable lesion than other traumatic intracranial hemorrhagic types, in case of timely surgery. Consistent with our results, EDH has generally been associated with favorable outcome (as opposed to other injury types) in cohorts of moderate to severe TBI and is considered beneficial, e.g., in the Rotterdam, Helsinki, and Stockholm radiological scoring systems [17, 29].

Not surprisingly, those with larger ASDH volumes more often required craniotomy, but, in similarity with larger EDHs, they were not more prone to require last-tier ICP treatments or develop more ICP insults. Patients with larger ASDHs were generally older (r = 0.41, p < 0.001) and therefore more likely tolerated larger hematomas due to certain brain atrophy, which could also explain why larger ASDH size did not independently translate into more ICP problems. Also, if brain swelling occurred during evacuation of the ASDH, the bone flap was usually not repositioned, an action that likely reduced postoperative ICP problems in many severe ASDH cases. Interestingly, larger ASDHs were independently associated with less episodes of low CPP, even after adjustment for age as a potential confounder for high ABP and low ICP. Furthermore, we only found a univariate association between larger ASDH size and higher PRx, but not in multiple linear regressions. The association was likely attenuated after adjustment by higher age which correlated with both high PRx and large ASDHs. ASDHs have been associated with higher PRx in some [34], but not all studies [35]. It has been suggested that acceleration/deceleration injury mechanism could be a driver of cerebral pressure autoregulatory disturbances (high PRx) [34, 35] and it is possible that differences in the underlying mechanism for the ASDH (elderly with falls vs. young patients in road traffic accidents) are crucial for the association with PRx. Large ASDHs are on the one hand extra-cerebral and potentially treatable if evacuated in time, but they are also associated with initiating underlying hemispheric injury including ischemia, reperfusion injury, and edema predisposing for secondary brain injury [21]. The combination of these factors likely explained its independent association with unfavorable outcome, consistent with previous studies [17, 29].

tSAH was common in the study population and occurred in 73% of the cases. tSAH typically includes a diffuse spread of small amounts of blood in the subarachnoid space not causing any mass effect, per se, which likely explained the lower rate of craniotomies in this group. However, still, tSAH correlated with the need to proceed with DC surgery and was also independently associated with more ICP problems. For example, tSAH has previously shown to be an indicator of evolving brain injury such as contusions that have not yet blossomed on early CT scans [2, 12]. Furthermore, there was no association between tSAH and PRx. This was to some extent surprising, since tSAH has been associated post-traumatic vasospasm [1] which likely overlaps with the pressure autoregulatory capacity to some extent. However, it is possible that our tSAH assessments (present or absent) were too crude to fully assess the burden of this lesion type. Currently, the literature is also conflicting, with studies both in favor [34] and against [35] an association between tSAH and high PRx. tSAH was also independently associated with unfavorable outcome. Although isolated tSAH in mild TBI cohort is a favorable finding, previous studies support that the opposite is true in moderate to severe TBI [14] and, e.g., tSAH carries a negative prognostic finding of worse outcome in the Rotterdam score [17]. In addition, IVHs were associated with a lower incidence of craniotomy and were independently associated with a lower percentage of ICP problems, possibly as it is rather a condition that is more effectively treated with an EVD in case cerebrospinal fluid circulation disturbances ensue.

Larger contusion volumes were associated with thiopental and DC treatment, but not with craniotomy. This was probably partly explained by the fact that many contusions were quite small (median 7 mL) without midline shift and also that total volume was based on the sum of potentially multiple, scattered contusions more resembling a diffuse injury. In addition, for the cases with more diffuse spread of contusions involving extensive areas of the brain, DC might have been preferred over evacuation [16]. Also, e.g., scattered multiple contusions with ICP problems without midline shift, thiopental might have been a feasible option for ICP control. Contusion volumes were not associated with cerebral pressure autoregulatory disturbances, consistent with some previous studies [34]. However, in more granular studies, Zeiler et al. established a potential link between autoregulatory disturbances and the risk of progression of peri-contusional edema [19, 20]. However, in this study, we assessed the total contusion volume and rather assessed PRx during the entire monitoring than between the radiological scans which could explain the discrepant findings. Lastly, contusions are per definition an intracerebral injury and greater size implies more extensive neurological injury. Consistently, greater contusion size was independently associated with worse outcome in this study.

Not surprisingly, indicators of significant mass effect, i.e., midline shift and compressions of basal cisterns, were associated with a worse clinical trajectory. Both correlated with a higher frequency of craniotomy and DC surgery. These measures of mass effect were not independently associated with more ICP insults, most likely since such malignant early signs of brain herniation required aggressive emergency neurosurgery (craniotomy and DC) which alleviated ICP post-operatively when it was monitored and analyzed. Still, compressed/obliterated basal cisterns were independently associated with disturbances in cerebral pressure autoregulation during the first 7 days post-injury, which was also found in a previous study [35]. This has been suggested to reflect severe diffuse brain injury caused by acceleration/deceleration forces [35] that may in turn increase the susceptibility for secondary injury due to cerebral ischemia and disturbances in energy metabolism [3, 24]. Obliterated basal cisterns were also independently associated with poor outcome, consistent with the Marshall classification and the Rotterdam and Helsinki scoring system [17, 18, 29]. The inclusion of this radiological measure of brain herniation in the multiple regression likely explained why the association between unreactive pupils and worse outcome was attenuated.

Furthermore, intracranial hemorrhage progression was independently associated with more ICP and CPP insults. This probably both reflected that these patients experienced greater intracranial hemorrhagic volumes leading to higher ICP, but also implied an evolving process between the CTs while ICP monitoring might have already been started and therefore captured in this data. However, hemorrhage progression and emergency neurosurgery were not independent predictors of worse outcome, as this was rather explained by specific hemorrhagic lesion types and mass effect.

On balance, classifications of intracranial hemorrhagic lesion types, sizes, mass effect, and evolution from early CT scans post-injury may be of value for predicting the clinical trajectories and development of secondary insults of important NIC variables such as ICP, CPP, and PRx. In future efforts, it may be of value to integrate such data into machine learning models for more granular predictions of these secondary insults during NIC [7] and long-term clinical outcome [8]. We also found that secondary ICP and CPP insults were rare, occurred in median in less than 5% of the monitoring time, and was overall only weakly associated with the majority of the intracranial hemorrhage types and their volumes. Thus, attentive NIC was important to control these physiological parameters under all circumstances, despite the great heterogeneity on early imaging in lesion types, volumes, and mass effect.

TBI is a heterogeneous disease with several different types of intracranial hemorrhagic lesions of various sizes, mass effect, and evolution. It is generally challenging to predict the clinical course and outcome. This study provides granular radiological analyses of lesion specific variables (such as lesion type, size, and progression) in relation to high-frequency variables of cerebral physiology (e.g., ICP burden and PRx status) and long-term clinical outcome. Future efforts should to integrate such granular data into more sophisticated machine learning models to aid the clinician to better anticipate emerging secondary insults and to predict clinical outcome.