External Lumbar Drainage for Refractory Intracranial Hypertension in Traumatic Brain Injury: A Systematic Review

Considerable variation exists in the clinical practice of cerebrospinal fluid diversion for medically refractory intracranial hypertension in patients with acute traumatic brain injury (TBI), which is achievable via lumbar or ventricular drainage. This systematic review sought to compile the available evidence for the efficacy and safety of the use of lumbar drains for intracranial pressure (ICP) control. A systematic review of the literature was performed with the search and data extraction performed by two reviewers independently in duplicate. Nine independent studies were identified, enrolling 230 patients, 159 with TBI. Efficacy for ICP control was observed across all studies, with immediate and sustained effect, reducing medical therapy requirements. Lumbar drainage with medical therapy appears effective when used alone and as an adjunct to ventricular drainage. Safety reporting varied in quality. Clinical or radiological incidents of cerebral herniation (with an unclear relationship to lumbar drainage) were observed in 14/230 patients resulting in one incident of morbidity without adverse patient outcome. The available data is generally poor in quality and volume, but supportive of the efficacy of lumbar drainage for ICP control. Few reports of adverse outcomes are suggestive of, but are insufficient to confirm, the safety of use in the appropriate patient and clinical setting. Further large prospective observational studies are required to generate sufficient support of an acceptable safety profile.


Introduction And Background
Established management practices for moderate and severe traumatic brain injury (TBI) are centred on minimising secondary injury through normalising intracranial homeostasis. This is achieved through a therapeutic paradigm principally centred on the avoidance of raised intracranial pressure (ICP) and maintenance of cerebral perfusion pressure (CPP). Contemporary ICP management strategies utilise a sequential escalation of therapeutic intensity until ICP control is achieved, with protocols based on the Brain Trauma Foundation (BTF) guidelines [1]. Initial medical treatments for intracranial hypertension include sedation, mild hypocapnia, and hyperosmolar therapy. Where intracranial hypertension is refractory to these interventions, therapies including diversion of cerebrospinal fluid (CSF), barbiturate coma, and decompressive craniectomy can be considered.
CSF diversion to manage raised ICP is classically understood in terms of the Monro-Kellie doctrine [2,3]. The principle states that diversion (or buffering) of one constituent of the intracranial compartment (e.g., CSF) permits an increasing volume of other constituents (i.e., parenchymal oedema in response to trauma). CSF can be removed from the system through two principal access points: (1) a ventriculostomy in the lateral ventricle connected to an external ventricular drain (EVD); or (2) a lumbar catheter connected to an external lumbar drain (ELD). CSF drainage can be achieved intermittently through discrete episodes of volumecontrolled drainage or continuously by permitting uninterrupted pressure-controlled drainage [4].
Traditionally, CSF diversion in TBI is achieved through a ventriculostomy. Siting a ventriculostomy also permits measurement of intraventricular pressure by transduction of a closed drain, though in recent years, ICP monitoring has more commonly been achieved through intraparenchymal monitoring via a transcranial access device or "bolt" [5]. Practice varies from centre to centre [6], and between adults and paediatrics, with the usage of EVD monitoring and drainage being a more common practice in paediatric trauma centres [7] but less so in adults [8].
ELD originated in 1963 as a means for reducing cerebral tension intraoperatively [9] and has become an established method of CSF diversion in a variety of settings [10], including post-traumatic CSF leak [11,12], normal pressure hydrocephalus assessment [13], skull base surgery [14,15], and in thoracoabdominal aortic surgery to reduce spinal cord ischaemia [16]. The use of ELD for ICP control after TBI is less common due to the potential risk of iatrogenic transtentorial herniation, recognised in historical examples of herniation after lumbar puncture in patients with raised ICP [17][18][19][20]. In other contexts of intracranial hypertension, the use of ELD has been successfully adopted, including in bacterial and cryptococcal meningitis [21,22] and subarachnoid haemorrhage (SAH) [23][24][25]; achieving similar ICP control to ventriculostomy drainage without significant complication rates, though consensus guidelines have found insufficient evidence to support ELD as an option in ICP control after TBI [26].

Rationale
Lumbar drainage represents a potential alternative means of CSF diversion to ventriculostomy due to avoidance of the passage of a drain through cerebral parenchyma. Insertion of a lumbar drain can be a technically and logistically simpler procedure, particularly in patients with isolated TBI with small lateral ventricles rendering insertion of EVD challenging or not achievable. Whilst thought to also achieve ICP control, concerns regarding the safety of lumbar drainage lie in the possible complication of cerebral herniation with associated morbidity and mortality. As such, to inform current practice and future research directions, there is a need for a summary of the safety and efficacy of external lumbar drainage for ICP control in TBI, particular with reference to the safety and efficacy of the other available CSF diversion modality, external ventricular drainage.

Objectives
Our objective is to systematically evaluate the available literature, examining the use of external lumbar drainage of CSF for refractory intracranial hypertension in traumatic brain injury (TBI). Specifically, we seek to summarise the available evidence for (1) the efficacy of the use of external lumbar drains in the control of refractory intracranial hypertension in TBI; and (2) the safety of external lumbar drainage in acute TBI.

Review Methods
A systematic review of the literature was performed following the methodology of the Cochrane Handbook for Systematic Reviewers and presented in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) [27]. The review was registered with PROSPERO under registration number CRD42020192283. No amendments were made to the information provided at registration or in the protocol. This article was previously posted to the ResearchSquare pre-print server on August 2nd 2021.
The primary review questions determined by the authors were: (1) Is CSF diversion via an external lumbar drain a safe procedure in TBI?; (2) Does CSF diversion via an external lumbar drain effectively control intracranial hypertension in TBI? Secondary review questions determined by the authors were: (3) What selection/exclusion criteria are utilised for patient selection to determine appropriate candidates for external lumbar drainage?; (4) Are there any established protocols for the use of external lumbar drainage for the control of ICP in TBI?

Inclusion Criteria
Population: The population of interest were human subjects with moderate or severe TBI, inclusive of both adult and paediatric populations.
Intervention: The intervention considered is CSF diversion via an ELD.
Outcome measures: The outcome measures of interest were ICP control, CPP optimisation and the safety profile of the intervention. A considered secondary outcome measure was any appropriate measure of functional outcome post-injury. Acceptable indices of ICP/CPP control considered are (1) direct measurement pre-and post-drainage, and (2) indirect outcome measures of reduced incidence of requiring osmotic therapy, hyperventilation, or more invasive surgical procedures (external ventricular drainage or decompressive craniectomy). For each outcome, any appropriate effect measure was accepted for the presentation of results; due to heterogeneity, no synthesis was performed and as such, no procedures were required for data conversion handling of missing summary statistics.
Setting: Given the intensive monitoring required to gain the above data, such studies would be in the intensive care unit or neurosciences critical care settings.
Methodology: Any research methodology in humans was considered for inclusion, including observational studies and case series/reports.

Exclusion Criteria
Studies conducted in animal models or in vitro human models were excluded. Studies which recorded the variables of interest, but did not report sufficient data on these variables for our research questions were also excluded.

Information Sources
We systematically searched the following database from their respective inception to September 2020: Medline, Embase, Cochrane Central, NICE Evidence, Google Scholar and Web of Science. Reference lists of pertinent review articles on the topic were hand-searched for suitable articles. Reference lists of identified articles for inclusion were also hand-searched for suitable articles. Appendix 1 provides the search strategies employed for each database.

Study Selection
Studies were independently screened for inclusion by two reviewers, utilising a referential standardised proforma. Eligibility for study inclusion was defined as: a study of human patients with TBI, with the measurement of intracranial pressure via intraparenchymal or intraventricular pressure monitor, with insertion of an external lumbar drain for the purposes of CSF drainage with the therapeutic aim of ICP/CPP control.

Data Extraction
Data were extracted from included reports by two reviewers: conducted independently in duplicate, using piloted forms. Data extracted included study characteristics, patient demographics, number of patients, ICP measurement method and frequency, external lumbar drainage timing and drainage protocol, patient selection/exclusion criteria, ICP/CPP control outcome, avoidance of treatment intensity level escalation, and functional outcome.

Risk of Bias in Individual Studies
Two reviewers assessed the risk of bias of each individual study using the Risk Of Bias in Non-Randomised Studies Tool (ROBINS-I) [28], rating each level of bias as low, medium, serious, or critical. Unresolved discrepancies were discussed with a third party.

Synthesis of Results
The heterogeneous nature of data collection and analysis methods precludes our ability to combine and synthesise results. As such, a descriptive summary will be presented, with all included studies tabulated. Data are presented as given by the respective authors: data was not converted to a homogenous outcome reporting metric. Any missing summary statistics are identified by stating their absence in the tabular results presentation. Qualitative assessment of the confidence in the body evidence for outcomes was performed and presented as a narrative.

Study Selection
A search of the literature pertaining to lumbar drainage in TBI yielded 610 results. After the removal of duplicates, 453 research items were screened against the inclusion criteria. Twenty-seven full text items were retrieved for eligibility assessment, of which 12 studies were considered to fulfil the criteria (Figure 1). Reasons for exclusion included commentary papers, review articles, and different procedures, further highlighted in the diagram.

Study and Patient Demographics
Three articles were presented as case series [29,32,34], with another presented as a case report and accompanying case series [37]. Five studies were presented as observational studies: three retrospective [31,33,35] and two prospective [23,25]. Three studies utilised a mixed cohort of patients with TBI and aneurysmal subarachnoid haemorrhage (SAH) and/or intracerebral haemorrhage (ICH) [23,25,35], with six studies including only patients with TBI [29,[31][32][33][34]37]. One study included only paediatric patients [32], seven studies included only adult patients [25,29,31,[33][34][35]37] and one included adults and children [23]. In total, 230 patients were included in the studies, of which 159 were patients with TBI. The demographics of included patients are reported in Table 1. Due to the manner in which the results were reported, the present authors were unable to delineate results pertaining only to TBI patients in the studies using a mixed cohort; as such the results will be presented together, noting where a mixed cohort is implicated where the results of these studies are discussed. Manet    Whilst our exclusion criteria resulted in the removal of studies which achieved CSF diversion through EVD alone, five of the nine studies included patients with an EVD in addition to ELD. In total, 100 patients received an ELD alone, and 130 patients received ELD + EVD. In three studies, all participants had both EVD and ELD [25,32,37], with EVD preceding ELD, two studies had a mixed cohort of ELD with and without EVD [23,35]. Llompart-Pou et al., 2011 [33] included two patients (of 30) who received an EVD though it is not clear whether the insertion of EVD preceded that of ELD. Patients in the remaining studies had ELD alone, with patients included in Manet et al., 2016 [34] only receiving ELD where EVD insertion had failed or was contraindicated. No study reported comparative outcome measures between groups with or without EVD. Furthermore, in such cases of ELD + EVD, there is variability of whether EVD was used for drainage, pressure monitoring, or both, with insufficient data reporting based on sub-group (ELD+EVD vs ELD) stratifications. As such, no conclusions may be drawn as to the relative potential advantages of implementing dual modality CSF drainage.
Eight studies only included patients with either: no surgical intracranial lesion or a surgical intracranial lesion post-evacuation; in one study, this was not specified [29]. All study protocols only instigated CSF diversion where medical management options had failed, all specifying at least hyperventilation and osmotic therapy in their ICP management protocols. Six studies indicated use of barbiturate coma [23,29,[32][33][34]37] and five indicated use of hypothermia [29,[33][34][35]37]. Four studies indicated the use of ventricular drainage prior to ELD [23,25,35,37] where possible.

Efficacy
All studies reported a marked reduction in ICP after CSF diversion with an ELD. Individual study results are reported in Table 2. The three studies [23,31,33] reporting the outcome of statistical tests of this effect all reported a statistically significant reduction in ICP after the introduction of lumbar drainage. Of four studies [23,25,29,32] presenting data or observations on the effect of lumbar drainage on CPP, all reported a positive effect. Four studies presented observations on the effect of lumbar drainage on treatment intensity or requirement for other ICP-lowering therapies [23,25,29,35]. All four reported a beneficial effect on the requirement of other therapies after the institution of lumbar drainage.

Safety
Of 230 patients, there were 14 reports of either radiological or clinical signs of cerebral herniation after insertion of ELD ( Table 3). One study [23] reported 12 of the 14 incidences of cerebral herniation, but the relationship between the development of clinical/radiological signs of cerebral herniation and the introduction of lumbar drainage is not described. The study did not present the relative incidence of these complications in TBI vs SAH patients, nor in ELD vs ELD + EVD patients. One of these episodes was reported as occurring due to the iatrogenic disconnection and uncontrolled drainage from the lumbar drain, and the patient survived without a neurological deficit. In the eight patients of these 12 who subsequently died, the authors did not attribute any death to, or as a complication of, lumbar drainage. There were two further incidences of herniation reported in the included literature, one patient exhibited only radiological signs [31] (bilateral uncal herniation), and one patient developed a fixed dilated pupil four hours after insertion of ELD which returned to normal after "an emergent craniectomy" and the patient made a "good neurological recovery" [25].   Seven out of nine CNS infection cases were reported to have both EVD and ELD in situ. There were 20 incidences of lumbar drain failure requiring revision. Eleven patients were reported to require long-term CSF diversion with the insertion of a CSF shunt.

CT Criteria
Seven of the nine studies stipulated the requirement of a CT head prior to proceeding with the insertion of an ELD [23,29,31,[33][34][35]37]. Two studies required this in the preceding 24 hours [29,33] and one in the preceding eight hours [31]. All studies requiring CT described imaging criteria [23,29,31,[33][34][35]37], all including discernible basal cisterns and the absence of a surgical mass/cerebral herniation. Three studies also included the requirement of a midline shift of <10mm [29,33,35]. Manet et al., 2017 [35] used the presence of "a gradual development of subdural or subarachnoid collections located in the Sylvian and/or interhemispheric fissures and/or cortical sulci" as evidence of external hydrocephalus and thus inclusion in their study. Bauer et al., 2017 [31] proposed and utilised a scoring system based on the patency of the prepontine and quadrigeminal cisterns and the absence of uncal and foraminal herniation on the preprocedural imaging.

ELD Drainage Protocol
Eight studies utilised an ELD in all patients [23,25,29,[32][33][34][35]37]. One study used a single lumbar puncture or intermittently open drainage in four of nine patients [31]. Of the eight studies which utilised only ELD, six used protocols with a continuous drainage strategy [23,29,[32][33][34][35]. One study [25] used varying methods, either continuous pressure-controlled drainage or fixed drainage at 10 ml/hr. One did not state their drainage protocol [37]. Four studies describe an initial high drainage protocol until ICP control was achieved, with intensive pupillary examination [23,32,34,35], two of which were in external hydrocephalus. Further details of specific drainage protocols are available in Appendix 2.

Heterogeneity of Studies
A descriptive review methodology was utilised due to the considerable heterogeneity of included studies. Practice variation in CSF diversion in TBI is considerable, relying on physician preference, availability of resources and differing guideline approaches for implementation of CSF diversion based on variable predefined ICP thresholds. Secondly, significant variation in of outcome measures in the included studies was identified. These included the modified Rankin scale, ICP metrics of efficacy (gross ICP or ICP change), and the Glasgow Outcome Scale. Similarly, complication reporting was variable in its detail and objectivity. Given such heterogeneity in both implementation of the intervention and in reporting of outcome measures, it was not deemed valuable or reliable to perform quantitative synthesis.

Risk of Bias Assessment
The risk of bias across was deemed low or moderate in all domains across all studies. Principle sources of moderate bias identified across studies were limited consideration of the possibility of confounding factors, limited consideration of selection bias, and classification of intervention. Overall, of the nine studies, all were deemed to have a moderate risk of bias (Appendix 3).

Discussion
The aim of this systematic review was to evaluate the available efficacy and safety data regarding the use of lumbar drainage for ICP control in TBI. The literature identified represents a small evidence base supporting the efficacy of lumbar drainage, where nine independent studies observed the effects in 159 patients with TBI (230 patients overall). Other than infection, need for drain revision, and transient mydriasis, one incident of morbidity (not affecting neurological outcome) was reported in these studies; though this is suggestive of the safety of ELD in TBI, the small cohort is insufficient to conclusively confirm it.

Efficacy
The data presented in the included studies supports the efficacy of lumbar drainage for ICP control as an adjunct to best medical therapy where optimal medical management has failed. Every study reported a beneficial effect for ICP control overall, with all of those performing statistical analysis finding a significant reduction in ICP following ELD insertion. Studies reporting mean reduction values utilised data one hour before and after commencing lumbar drainage, which unanimously confirmed an immediate effect. The positive effects on reducing the need for medical interventions such as osmotic therapy are also encouraging and were observed in all four studies reporting this outcome. The effect has been observed, though only reported for small numbers of patients, with increasing efficacy up to three days after the instigation of lumbar drainage [29]. Together with one-hour pre-and post-ELD ICP readings, the effect of lumbar drainage appears to be both immediate and sustained.
A significant reduction in ICP was reported in four studies which included ventriculostomy drainage prior to lumbar drainage [23,25,35,37], suggesting that ELD offers both a benefit when used as the sole modality of CSF diversion and an additional benefit when used in combination with EVD.
Functional outcomes were not considered as a marker of "efficacy"; rather, ICP control can be considered as a surrogate marker of potentially improving functional outcomes due to the association of intracranial hypertension and poor outcome [38]. The numbers enrolled in these combined studies are by far insufficiently powered to assess the effect on functional outcome.

Safety
The available data identified by this systematic review is insufficient to draw firm conclusions on the safety profile of ELD in the management of raised ICP in TBI; however, the included studies here have not reported incidents of the usage of ELD directly contributing to mortality. A single case of ELD usage has been reported to result in pupillary changes requiring urgent surgery with good neurological recovery). However, the absence of significant complications in a series of 159 patients with TBI is unlikely to generate sufficient confidence to assuage the reservations held by some neurosurgical centres. In the single reported case of cerebral herniation requiring surgery, this was urgently recognised and appropriately managed, resulting in good neurological recovery. Despite historical concerns of "coning" from lumbar puncture or drainage, the results suggest that in the modern intensive care unit, there is sufficient clinical monitoring to facilitate early recognition of cerebral herniation through clinical and radiological assessment and prompt remedial action.
Whilst a high mortality rate is not justification for therapeutics with the potential for harm, the alternative to lumbar drain is typically insertion of EVD; where this fails, decompressive craniectomy is typically indicated, and neither can be considered to be low morbidity. Whilst not robust, the results here are suggestive of a low to very low risk of severe morbidity or mortality as a result of ELD in selected patients. When considering surgical risk, ELD offers the potential benefit of achieving ICP control without the need for EVD or decompressive craniectomy.
Infection rates of the included studies were 3.9%, though some incidences of infection were in patients with both EVD and ELD. This incidence is similar to that reported in the literature, with infection rates of ELD reported as 4.2-7% [39,40] and EVD reported as 7-9% [39,41].

CT Criteria
Given the concern of obliterated basal cisterns preventing CSF movement across the foramen magnum, The absence of tonsillar herniation, midline shift of less than 10 mm, and patency of prepontine and interpeduncular cisterns have been more recent additions. Due to the low complication rate, it would be difficult to identify which protocol is optimal for safety in patient selection. Further details of the radiological features of those with transient mydriasis are not available to make a post-hoc assessment of which further features transpired to be associated with risk.

Drainage Strategies
Drainage strategies varied amongst studies, with insufficient data reporting to draw conclusions regarding the optimal protocol. The protocols where described adopt broadly similar strategies: greater drainage permitted to achieve ICP control (<20 mmHg), with subsequent "maintenance" drainage (ICP in the region of 10-20 mmHg) and closure of the drain where ICP was low (<10 mmHg). Pressure settings of the ELD varied widely between 0 cm above the foramen of Monro (FOM) and 20 cm above. The study reporting a fixed dilated pupil four hours after ELD insertion allowed free drainage at 0 cm [25], a strategy which physiologically would permit continued drainage in the context of early cerebral herniation. Similarly, Tuettenberg et al., observing 12 episodes of unilateral mydriasis in their mixed TBI/SAH cohort, permitted free drainage at 5 cm above the FOM. Differing strategies for drainage should also consider nursing practicalities: continuous drainage at lower pressure settings is more likely to result in high volume outputs between nursing monitoring intervals.
When considering the physical and biomechanical aspects of drainage, a minimum volume is required to produce an acceptable ICP/CPP for the purpose of optimising perfusion and minimising the requirement for intervention. Therefore a sliding scale hourly drainage target, with an arbitrary maximum per hour of 10 ml, may provide an attractive and safe starting point in the views of the authors. Physically the volume of any potentially herniated neuro-axial tissue cannot exceed that of the volume of fluid drained, supporting further the likely safety of small incremental drainage as opposed to open pressure regulated drainage where theoretically no limit on the volume of drained fluid or herniating tissue exists.

Concomitant EVD
As outlined above, ELD appears to offer additional benefits when used as an adjunct to EVD in ICP control.
In the largest included study [23], 84/100 patients had received ventricular drainage prior to ELD which had failed to control their ICP, though it is not specified in the methods whether ventricular drainage continued once ELD was inserted. Given that the concerns surrounding lumbar drainage are centred on a pressure gradient across the foramen magnum, simultaneous EVD and ELD drainage theoretically reduces the risk of this occurring. However, this study reported a relatively high rate of signs of cerebral herniation (with unclear timing in relation to ELD).

Comparison of CSF Diversion Techniques
One important consideration is the relative complication rate of the two strategies: EVD and ELD. A systematic review [42] found the main (early) EVD-related complications included: infection (12.8%), haemorrhage (1.5%), device failure (14.9%), and malposition (10.1%). This review focused on the timing of intervention and identified that early EVD intervention correlated with a higher infection rate (12.8%) than late (8.3%). Indirect comparison with ELD complication rates identified in this study demonstrates broadly similar rates. Included studies reported infection associated with ELD (5.7%) and device blockage (17.0%) complications. Three included studies did not report any complications from ELD use in their cohorts [20,22,37]. Given the predominance of retrospective study designs, underreporting of device-associated complications is likely, and the risk of reporting and observation is high.
The prospect of utilising ELD instead of, or as an adjunct to, EVD in selected patients has some potential benefits. The prevalence of "slit" ventricles in the TBI cohort with intracranial hypertension presents a technical challenge. The principal consideration of the use of EVD or ELD in the context of TBI is overall safety. The included studies have demonstrated that, in particular, in patients with stringently controlled drainage and monitoring parameters, ELD can be implemented safely in this context. However, further prospective study is required to be able to draw valid conclusions on the relative risks and merits of the two available modalities.

Limitations
The limitations of the present study are primarily due to the heterogeneity of both the included studies and the data that they have presented. Exclusion of studies with a mixed cohort of brain injury aetiology and adult/paediatric cohort would have further narrowed the low patient and study numbers available for consideration by this study. As such, we have included these, particularly given the necessity to be comprehensive when considering the intervention's safety profile. Further to this, even with these inclusions, the numbers of patients and studies are low, though identifying paucity in the literature serves as an outcome of this review. The results should be interpreted with caution in recognition that the available literature is a severe limitation of this study. However, this serves to highlight the necessity for high-quality evidence through future studies. The detail in reporting the potential complication events was variable but generally poor. This is, in part, likely to be related to the incidence of clinical and radiological evidence of "coning" being a potential part of the natural history of severe TBI with refractory intracranial hypertension. As such, we have only considered complications to be related to lumbar drainage, where the respective authors have themselves reported these events to be a consequence of lumbar drainage. Given the heterogeneity of cohorts and data reporting, it was not deemed suitable nor beneficial to perform a meta-analysis, and a narrative review of the data is inherently limited in the conclusions that it may draw.

Conclusions
The literature appears to support the efficacy of lumbar drainage in ICP control where medical management has failed, though the data on its safety is encouraging but insufficient to draw conclusive recommendations. The efficacy and safety profile of ELD in comparison to EVD is not known, though ELD appears to be beneficial for ICP control both alone and where EVD has failed. Although the evidence base is insufficient to draw firm conclusions, based on the available evidence, there is no clear indication that the complication rates of ELD are greater than those of EVD. Further large prospective observational studies are required to generate sufficient support for an acceptable safety profile, with the possibility of subsequent randomised controlled studies to ultimately assess therapeutic parity.

Appendix 1
The following search strategies were employed from these databases:

Conflicts of interest:
In compliance with the ICMJE uniform disclosure form, all authors declare the following: Payment/services info: All authors have declared that no financial support was received from any organization for the submitted work. Financial relationships: All authors have declared that they have no financial relationships at present or within the previous three years with any organizations that might have an interest in the submitted work. Other relationships: All authors have declared that there are no other relationships or activities that could appear to have influenced the submitted work.