Neuromonitoring is used to guide treatment in patients with acute brain injuries. Most neuroscience intensive care units (NICU) in high-income countries have intracranial pressure (ICP) and cerebral perfusion pressure (CPP), along with transcranial Doppler (TCD) and surface electroencephalography (sEEG) as brain monitoring tools available in a selection of their acute brain injured patients (Le Roux et al., 2014; Hutchinson et al., 2015; Carney et al., 2017; Cnossen et al., 2017). Partial pressure of brain tissue oxygenation (PbtO₂), cerebral temperature (Cerebral T), regional cerebral blood flow (rCBF), jugular bulb venous oximetry (SvjO₂), cerebral microdialysis (CMD), near-infrared spectroscopy (NIRS) and electrocorticography (ECoG; from invasive electrodes on the cerebral surface) and depth electroencephalography (dEEG) are the other frequently applied modalities (Le Roux et al., 2014; Stocchetti et al., 2017).

Cerebral multimodality monitoring (MMM) is often mentioned in NICU reviews (Makarenko et al., 2016; Stocchetti et al., 2017; Tasneem et al., 2017; Smith, 2018; Al-Mufti et al., 2019; Veldeman et al., 2020a; Yang, 2020), but reviews and guidelines mainly discuss the results of unimodal signals (Le Roux et al., 2014; Carney et al., 2017). The practical application of “combining modalities” is limited by the high-dimensionality of signals and non-standardized methods to present the information at the bedside. Also, clinical context, including imaging results, is not incorporated (Tasneem et al., 2017; Smith, 2018; Al-Mufti et al., 2019; Veldeman et al., 2020a; Yang, 2020). In 2014, Le Roux et al. (2014) formulated five-year expectations and recommendations regarding MMM in acute brain injured patients. They expected patient-specific rather than population-specific thresholds, TCD-based non-invasive measures for ICP monitoring, and advances in the detection of cortical spreading depolarization.

Since the projections by Le Roux et al. were put forward, no overview of the application of MMM studies has been published (Le Roux et al., 2014). However, rigorous insight into MMM of recent years could detect benefits, pitfalls, and gaps for improving future clinical study designs. In this narrative review, we, therefore, aim to investigate the recent applications of cerebral MMM in studies for acute brain injured patients (i.e., adult patients with traumatic brain injury (TBI), subarachnoid hemorrhage (SAH), intracerebral hemorrhage (ICH), acute ischemic stroke (AIS) or hypoxic ischemic brain injury following cardiac arrest (HIBI). Our objectives are (I) to identify which combinations of monitoring modalities are currently applied, in general, and across the different acute brain injuries, (II) to summarize the monitoring setting, study setting, and clinical characteristics, and (III) to discuss the potential added value of MMM on clinical outcome in intervention studies.

We identified studies describing combinations of cerebral monitors providing data that updates continuously or on a regular daily basis (i.e. regularly over the day) through a PubMed literature search. We used a stepwise approach for the literature search and identification of eligible studies.

Step 1, for each cerebral monitoring modality, a single PubMed query was used (Supplementary Table S1).

For objective I, we described the number and combinations of the different modalities. The number of monitoring combinations was calculated, and their synergy was visualized in a Circos plot (Krzywinski et al., 2009).

For objective II, we summarized the monitoring setting, study setting, and clinical characteristics of the selected studies between the diseases and reported the results as frequencies or medians (together with interquartile range, q1-q3). Furthermore, we described the secondary injuries studied in observational and interventional studies. Finally, we summarized the interventions that were applied in the MMM studies.

For objective III, we discussed the added value of MMM on clinical outcome in intervention studies.

After the abstract, references, and citation identification, 209 full-text studies were read. From these, 97 studies whose aim or objective(s) were not related to MMM were excluded. These excluded studies were predominantly (52 studies) validation (superiority/inferiority) studies comparing non-invasive TCD-based ICP with invasive ICP monitoring (25 studies). Supplementary Table S3 lists the modalities used for validation. The study selection flowchart is shown in Supplementary Figure S1. In addition, the number of included studies by year can be found in Supplementary Figure S2. For the final analysis, 112 MMM studies were available, of which 59 concerned TBI (53%), 53 SAH (47%), 13 ICH (12%), 5 AIS (4.5%), and 9 HIBI (8%).

We identified 11 monitoring modalities that update continuously or on a regular daily basis. The anatomical locations are graphically presented in Figure 1, showing eight invasive (ICP, PbtO₂, Cerebral T, rCBF, SvjO₂, CMD, ECoG, dEEG) and three non-invasive (TCD, NIRS, sEEG) modalities. The synergy of the combinations is shown in Figure 2. The individual modalities were integrated into 47 unique combinations (Figure 3). In 58 studies (52%), two modalities were applied, three in 28 studies (25%), and only 26 studies (23%) utilized more than three modalities (Supplementary Figure S3). ICP monitoring was the most frequently combined modality, in 92 studies (82%), with the highest number in TBI patients (53 studies, 90%). The second most applied modality was PbtO₂ in 71 studies (63%). SvjO₂ monitoring was only applied in six studies (5.4%) and mainly combined with ICP (5 studies) and PbtO₂ (5 studies) monitoring. Invasive neuronal activity monitoring (ECoG and dEEG studies, 17 studies) was more common than non-invasive neuronal activity monitoring (sEEG, 10 studies). Regarding non-invasive modalities, TCD was most often studied (25 studies), predominantly in patients with SAH, ICH, and AIS. TCD was not studied in HIBI patients. We studied only modalities that were part of the research protocol. However, 21 SAH studies also mentioned other modalities (mainly ICP, Cerebral T, and TCD), which were only part of the clinical protocol. These modalities were not considered as often only limited, or no continuous information was provided. Supplementary Table S5 lists these modalities for the individual studies. Lastly, only 58% of the studies analyzed more than 24 h of data per patient. A summary of the monitoring settings is given in Table 1 and Supplementary Table S4A.

The study setting and clinical characteristics are summarized in Table 2 and Supplementary Table S4B. Most were single-center studies (90 studies, 80%) with a median sample size of 36 (q1-q3, 20–74) patients. In 34 studies (30%), patients were enrolled over a period of more than 5 years. TBI studies included more patients compared to SAH studies (TBI 43 (22–100) patients versus SAH 26 (17–69) patients). In addition, TBI studies more often had a multicenter design (TBI, 37% versus SAH, 15%).

Eighteen studies (16%) included combinations of acute brain injured patients. Especially, ICH and AIS were combined with other acute brain injuries. There were only four single disease studies of ICH and only three of AIS. Although HIBI is the least contributing group, relatively more single disease studies were included (5 studies) compared to ICH and AIS.

Clinical characteristics differed between diseases. TBI studies included relatively younger male patients (71%<50 years, 75% male), whereas SAH studies included older female (70% 50–59 years, 31% male) patients. HIBI studies included middle-aged, slightly more male patients (67%, 40–49 years, 61% male). Studies that included ICH patients included a wide range of ages (40–69 years, 53% male). AIS included predominantly patients within the range 50–59 years and female (39% male).

Secondary brain injuries are heterogeneous in presentation, with a complex interplay between impairments in diffusion, perfusion, metabolic derangements, and neuronal damage. We studied the different conditions and phenomena defined by the authors of the observational (68 studies) and interventional (44 studies) studies. Authors reported hypo-/hyper perfusion, cerebrovascular autoregulation impairment, ICP plateau waves, spreading depolarization, diffuse cerebral ischemia, vasospasm, and metabolic distress. Due to the inconsistencies in definitions and nomenclature of (single) modalities, no detailed group results across the diseases are presented, but examples are given to explain these inconsistencies.

Authors either allocated patients with/without a specific secondary brain injury and compared differences in MMM signals between the groups, or authors selected a whole group of a particular disease. Then, they reported the secondary brain injuries based on the thresholds of each modality.

In general, the number of secondary brain injuries is large because each modality has its own threshold for impairment, or a combination of modalities defines an impairment. In other words, the definitions of secondary brain injuries are limited by the number of available modalities. For example, Lindner et al. (2021) defined mitochondrial dysfunction (single modality) as: CMD lactate/pyruvate (L/P)-ratio ≥ 40 + CMD-pyruvate ≥ 70 μmol/L, whereas Khellaf et al. defined mitochondrial dysfunction (three modalities) as: CMD L/P-ratio>25 for more than 2 h, ICP <20 mmHg; PbtO₂ <15 mmHg; PRx <0.3; brain extracellular glucose >1 mmol/L (Khellaf et al., 2022).

In addition, there were inconsistencies in nomenclature for impairments using single modalities. For example, Hosmann et al. (2022) define indications for cerebral ischemia as CMD L/P-ratio >40 CMD-glycerol >100 μmol/L, CMD-lactate >4 mmol/L, whereas Nyholm et al. (2017) defined cerebral ischemia as CMD-L/P ratio >40 and CMD-pyruvate <50 mol/L). For brain tissue hypoxia monitored by PbtO₂ there were in general two definitions used: PbtO₂ <15 mmHg (Burnol et al., 2021; Hosmann et al., 2021) or <20 mmHg (Le Roux et al., 2014; Gagnon et al., 2020; Sekhon et al., 2020; Gouvea Bogossian et al., 2021).

We identified systemic- (24 studies), cerebral (10 studies) interventions, and interventions guided by MMM (11 studies). Table 3 and Supplementary Table S4C summarize the study classifications. In addition, one study was classified as MMM-guided and a cerebral intervention.

A total number of 20 different systemic- or cerebral interventions were applied. An example of a systemic intervention is the administration of red blood cell (RBC) transfusion (Sekhon et al., 2015; Kurtz et al., 2016; McCredie et al., 2017; Gouvêa Bogossian et al., 2022). An example of a cerebral intervention is the application of prostacyclin with a beneficial effect on neuronal cell membrane destruction (Koskinen et al., 2019). Examples of MMM-guided interventions are the studies of Veldeman et al. They evaluated outcome between periods before and after introducing an invasive MMM-guided protocol to avoid PbtO₂ < 10 mmHg and CMD L/P-ratio > 40 in severe SAH patients with suspicion of delayed cerebral ischemia (Veldeman et al., 2020a; Veldeman et al., 2020b).

Interventions in the MMM studies serve mainly three purposes. Firstly, monitoring the effectiveness of an intervention. Secondly, collecting monitoring data in combination with an intervention for outcome evaluation/prediction. A third purpose is monitoring the need for an intervention. In other words, interventions guided by MMM to investigate the interplay between monitoring and a combination of (in general, systemic) interventions. Figure 4 illustrates the purposes of the interventions across the MMM studies.

To give insight into the range of systemic-, cerebral-, and MMM-guided interventions, we classified them into nine categories: ABP management, biomarkers, fluid management, mixed (combination of different) interventions, RBC-transfusion, physical (movement) interventions, vasospasm therapy, ventilation management, and other interventions. The number of studies per group is mostly less than five. The largest groups are the mixed interventions used to guide MMM (11 studies), followed by ventilation management interventions (10 studies). On the other hand, biomarkers and physical (movement) interventions were studied in only three studies. The specific interventions and the corresponding number of studies per category are shown in Table 4.

Clinical outcome is a study endpoint in 18 (41%) of the 44 interventional studies. Systemic- and cerebral interventions evaluated MMM for either outcome prediction (1 study) (Lubillo et al., 2018) or monitoring the effectiveness of an intervention in both the MMM signals and clinical outcome (8 studies) (Hockel et al., 2016; Jakkula et al., 2018; Ding et al., 2019; Sekhon et al., 2019; Svedung Wettervik et al., 2020b; Dagod et al., 2021; Kovacs et al., 2021). For MMM-guided intervention studies, clinical outcome resulted from the interplay between MMM and interventions. Nine (9/11, 82%) of the MMM-guided included clinical outcome as an endpoint, of which seven showed an improved outcome in favor of the MMM-guided group (78%, 7/9 studies). Five (45%) studied an ICP and PbtO₂-guided treatment in either TBI (Lin et al., 2015; Okonkwo et al., 2017; Sekhon et al., 2017) or SAH (Rass et al., 2019; Gouvea Bogossian et al., 2021) patients. Two of these compared pre-/post implementation of an MMM-guided protocol. Okonkwo et al. (2017) studied the feasibility and safety of an ICP and PbtO₂ protocol in a randomized controlled trial (RCT). Their study showed lower mortality and improved outcome, but the effects did not reach statistical significance. This was attributed to the small sample size. In addition, Rass et al. (2019) studied the brain hypoxia burden in two centers and found no difference in PbtO₂-levels and clinical outcome. The remaining four MMM-guided studies that showed an improved clinical outcome included the following modalities: (I) CMD in combination with ICP, PbtO_2, TCD (Veldeman et al., 2020a; Veldeman et al., 2020b) (II) ICP, PbtO₂, Cerebral T, and SvjO₂ (Fergusson et al., 2021), and (III) ICP, PbtO_2, rCBF, and TCD (Bele et al., 2015).

The principal insights gained from our analysis of the MMM literature are that: (Insight I) most reports of MMM involve just two monitoring modalities, one of which is typically ICP monitoring; (Insight II) we found relatively often 10 (8.9%) ECoG and 7 (6.3%) dEEG studies, of which 8 (50%) investigated cortical spreading depolarization; (Insight III) our results show that MMM is primarily used in TBI and SAH patients. In addition, ICH and AIS are sparsely studied as a single study population but mainly combined with other acute brain injuries. One of the reasons could be that (non) invasive cerebral monitoring was not part of HIBI, AIS, and ICH (international) treatment guidelines and protocols compared to TBI and SAH patients; (Insight IV) most MMM studies had an observational design without direct clinical and therapeutic implications at the bedside; (Insight V) The sample sizes are in general small with long inclusion periods; (Insight VI) a large variety of interventions were studied in limited numbers of studies; (Insight VII) seven of the nine MMM-guided intervention studies showed a significant improved clinical outcome in favor of treatment guided by MMM.

Cerebral MMM in neurocritical care patients with acute brain injury focuses predominantly on bimodal monitoring, studied mainly in TBI and SAH patients. Definitions of secondary injuries are limited by the number of modalities and differ in entity due to different thresholds. In addition, the applied interventions are large in variety, but they are limited in the number of studies. Although the improved clinical outcome in MMM-guided intervention studies strengthens the belief in this application, further interdisciplinary collaborations are needed to overcome the heterogeneity. Future research should focus on improved data collection, sample sizes, refining definitions of secondary injuries, and standardized interventions. Only then can we proceed with complex outcome studies with MMM-guided treatment.