Neurocritical Care

, Volume 14, Issue 1, pp 68–72

Continuous and Intermittent CSF Diversion after Subarachnoid Hemorrhage: a Pilot Study


  • G. S. Kim
    • Department of NeurologyNational Health Insurance Corporation Ilsan Hospital
  • A. Amato
    • Neuroscience Critical Care UnitDuke University Medical Center
  • M. L. James
    • Neuroscience Critical Care UnitDuke University Medical Center
    • Department of AnesthesiologyDuke University Medical Center
    • Department of Medicine (Neurology)Duke University Medical Center
  • G. W. Britz
    • Department of Surgery (Neurosurgery)Duke University Medical Center
  • A. Zomorodi
    • Department of Surgery (Neurosurgery)Duke University Medical Center
  • C. Graffagnino
    • Neuroscience Critical Care UnitDuke University Medical Center
    • Department of Medicine (Neurology)Duke University Medical Center
  • M. Zomorodi
    • Neuroscience Critical Care UnitDuke University Medical Center
    • University of North Carolina at Chapel Hill School of Nursing
    • Neuroscience Critical Care UnitDuke University Medical Center
    • Department of Medicine (Neurology)Duke University Medical Center
Original Article

DOI: 10.1007/s12028-010-9401-y

Cite this article as:
Kim, G.S., Amato, A., James, M.L. et al. Neurocrit Care (2011) 14: 68. doi:10.1007/s12028-010-9401-y



We examine two accepted methods of managing cerebrospinal fluid (CSF) drainage in patients following subarachnoid hemorrhage (SAH). The first is intermittent CSF drainage when intracranial pressure (ICP) reaches a pre-defined threshold (monitor-first) and the second is continuous CSF drainage (drain-first) at set pressure thresholds. This pilot study is designed to determine if there is a cause for a randomized study of comparing the two methods.


This prospective observational pilot study enrolled 37 patients with SAH and external ventricular drainage between October 2008 and August 2009. Patients were treated with one of two methods of ICP management (drain-first vs. monitor-first) according to the discretion of the admitting physician.


There were no significant differences in baseline characteristics including age, gender, severity of neurological dysfunction, and radiographic findings between the two groups. The incidence of vasospasm was not different between the drain-first group (66.7%; 16 of 24 patients) and the monitor-first group (53.9%; 7 of 13 patients).


This pilot study was neither powered, nor expected to detect a difference between groups. The results of this study provide support for the design and conduct of a randomized study to assess the impact of two methods of CSF diversion for patients with SAH.


Cerebral vasospasmComplicationExternal ventricular drainageIntracranial pressureSubarachnoid hemorrhage


Ruptured cerebral aneurysms result in a subarachnoid hemorrhage (SAH) and may lead to cerebral artery vasospasm in up to 50% of patients. The resulting vascular constriction, if left untreated, can cause delayed ischemic injury to the brain [1]. Ischemia associated cerebral edema may increase intracranial pressure (ICP) and lead to further ischemia and infarction [13].

The etiology and pathogenesis of vasospasm after SAH are not completely understood. However, because the time period during which erythrocytes break down and blood disappears from the cerebrospinal fluid (CSF) closely mirrors the onset and resolution of clinical vasospasm, it is widely accepted that blood products, such as oxyhemoglobin, act as spasmogens and contribute to vasospasm [46]. A high correlation also exists between the severity of the vasospasm and the volume of subarachnoid blood in the cisterns and ventricular system [6, 7]. It has been postulated that decreasing the burden of subarachnoid blood may reduce the likelihood of developing vasospasm. If this postulate is confirmed, CSF drainage may be an effective method of reducing vasospasm.

Intraventricular catheters are often inserted in patients with moderate to severe degrees of SAH in order to monitor ICP, allow for CSF diversion and treat acute hydrocephalus [6]. Duke University currently supports two different methods of CSF drainage and ICP monitoring for patients diagnosed with SAH. The first method is through intermittent CSF drainage with a focus on monitoring the ICP (monitor-first); the second is through near-continuous CSF drainage at set pressure thresholds with intermittent monitoring of ICP (drain-first). Continuous CSF drainage theoretically may increase the volume of bloody CSF output, thus reducing the likelihood of developing cerebral vasospasm. A potential disadvantage of this strategy is ventricular collapse from aggressive CSF drainage; the reduced cisternal subarachnoid space will then prevent CSF flow and retain spasmogens in the adjacent vessels. Intermittent CSF drainage may resolve this dilemma by allowing the removal of spasmogens while still permitting the flow of CSF through normal pathways during the interval between external drainage and ICP monitoring [811].

The purpose of this pilot study was to explore the safety and feasibility of testing the hypothesis that the method of CSF drainage impacts the incidence of cerebral artery vasospasm and to provide data for developing a randomized clinical trail of this hypothesis.

Materials and Methods

Patient Population

This pilot study was conducted using a prospective, observational comparative design. After receiving institutional review board (IRB) approval, subjects were recruited from a 16 bed Neurocritical Care Unit (NCCU). In all cases, consent was obtained from the patient or patient’s legally authorized representative prior to enrollment. One-hundred consecutive SAH patients were screened with 37 patients treated with external ventricular drainages (EVDs) between October 2008 and August 2009 enrolled in this study. Exclusion criteria included age younger than 18 years, Glasgow coma scale <4, or prisoners. Reasons for screen failure included lack of EVD treatment (N = 50), unable to obtain informed consent (N = 12), and age less than 18 years (N = 1). A complete record of examinations, clinical course, and procedures associated with EVD was prospectively maintained for each patient while duration of the patient’s stay in the NCCU was captured through chart abstraction from the electronic healthcare record.

Management Protocol

All patients received standard management of treatment for SAH including admission to NCCU, hemodynamic monitoring, and ICP management. Early aneurysm repair with surgical clipping or endovascular coiling was used at the discretion of the attending neurosurgeon (A.Z., or G.B.). Per standard of care at our institution, EVD placement was performed for patients with GCS <9 or those who were given a clinical diagnosis of acute hydrocephalus. At the discretion of the attending neurosurgeon, patients were treated with one of the two methods of EVD management that are standard-of-care in our NCCU. The duration of EVD was determined by the attending neurosurgeon.

Patients in the monitor-first group received continuous ICP monitoring and intermittent CSF drainage. Intermittent drainage was accomplished by setting the EVD system to a physician prescribed threshold and opening the stopcock to allow CSF drainage if and only if the ICP exceeded that threshold for a specified period of time. An example of this order would read: “Open EVD and drain CSF when ICP exceeds 20 mm Hg for 5 min.”

Patients in the drain-first group received continuous CSF drainage with intermittent ICP monitoring. Continuous drainage is accomplished by setting the EVD system to a physician prescribed threshold and keeping the stopcock open to allow for CSF drainage. When continuous CSF drainage is ordered, the stopcock from the EVD remains open and ICP is assessed at pre-determined intervals. An example of this order would read “Leave EVD open to drain at 10 mm Hg and monitor ICP once each hour.”

All patients received standardized serial neurological examinations performed until discharge. Transcranial Doppler (TCD) was performed daily by a technician each morning for the first 14 days following SAH. Moderate vasospasm was defined as a mean flow velocity in any insonated vessel with a TCD reading of >150 cm/s. When new neurological deficits were noted on physical examination or vasospasm was detected by TCD, a non-contrast head CT was ordered. At the discretion of the medical team, a CT-Angiogram (CTA) or catheter cerebral angiogram was performed to further evaluate for the presence of vasospasm. Angiographic spasm was defined as severe arterial narrowing (>50% stenosis) on CTA and catheter cerebral angiogram [12]. Patients with symptomatic cerebral vasospasm were initially treated with induced hypertension and hypervolemia. Target blood pressure and fluid volume were individualized to each patient. Hypertension was generally directed as a systolic arterial blood pressure >180 mmHg, and hypervolemia was determined by a combination of fluid volume status, daily weight, central venous pressure, and stroke volume variation (LiDCO, Covidien, London, UK), with careful observation and treatment for complications related to hyperdynamic therapy.

Statistical Analyses

All data were recorded and stored in a confidential manner in a computer spreadsheet format (Microsoft Excel™, Mountain View, CA). Where applicable, results were recorded as means ± standard error of the mean. The chi-square or Fisher exact tests were used for univariate analysis of categorical variables, and the t test was used for univariate analysis of continuous variables. Statistical analyses were performed using commercially available software (JMP 8.0, SAS Institute, Cary, NC). Significance was set at a probability value lower than 0.05.


Thirty-seven patients were enrolled and completed the protocol. There were no significant differences in baseline characteristics of age, gender, race, Hunt & Hess grade, Fisher grade, and methods of aneurismal repair was observed between the drain-first group and the monitor-first group (Table 1).
Table 1

Demographics and characteristics


Drain-first group (N = 24)

Monitor-first group (N = 13)

p value

Mean age in years

53.1 ± 2.43

55.9 ± 3.31


Female gender

16 (66.7%)

9 (69.2%)









 African American








Median Fisher grade




Median Hunt & Hess grade




Treated with coil

9 (37.5%)

4 (30.8%)


Treated with clip

10 (41.7%)

4 (30.8%)


The primary outcome, vasospasm, was scored dichotomously as present or absent and determined using a restricted conservative definition (angiogram as gold standard) and unrestricted liberal definition wherein vasospasm present by any diagnostic tool (CTA, TCD, angiogram) resulted in a score of present (Table 2). Vasospasm evidenced by TCD or radiological studies was identified in 23 patients (62.2%). There was no significant difference in the overall incidence of vasospasm between the drain-first group (66.7%; 16 of 24 patients) and the monitor-first group (53.9%; 7 of 13 patients). The incidence of vasospasm based on TCD, CTA, and cerebral angiogram (drain-first group; 62.5, 16.7, 16.7%, respectively; monitor-first group; 53.9, 23.1, 15.4%, respectively) showed no difference.
Table 2

Clinical outcomes


Drain-first group

Monitor-first group

p value

Mean highest ICP value (mmHg)

49.9 ± 4.82

43.9 ± 6.54


Mean CSF output per day (ml)

134 ± 17.1

135 ± 22.8


Mean EVD dwell time (days)

15.0 ± 1.15

13.4 ± 1.56


Mean length of ICU stay (days)

21.3 ± 0.24

18.9 ± 2.67


Mean modified Rankin score at discharge

3.75 ± 0.35

3.61 ± 0.48


Vasospasm by any diagnostic criteria

16 (66.7%)

7 (53.9%)


Shunt placed

7 (29.2%)

1 (7.7%)


EVD-related complicationsa

14 (58.3%)

4 (30.8%)



3 (12.5%)



 Clogged or blocked EVDb

10 (41.7%)

3 (23.1%)


 Self-removal of EVD

1 (4.2%)

1 (7.7%)


aComplications recorded include ventriculitis, clogged or blocked EVD, and self-device removal

bAt least one recorded episode

Secondary outcomes associated with ICP monitoring and CSF drainage were explored for both groups and there was no difference in the highest ICP value, CSF output per day, length of ICU stay, and modified Rankin score at discharge between two groups (Table 2). Although our small sample sized precluded reaching statistical significance, the drain-first group demonstrated a trend towards higher shunt rate at hospital discharge and more EVD complications (ventriculitis, clogged/blocked catheter, and self-removal of EVD) compared to the monitor-first group (Table 2). To further explore these variables given the small sample size, we calculated the odds of observing the condition given the treatment option (OR). The OR for complications was lower (0.21) in the monitor-first group (95% CI 0.05–0.98) than the drain-first group (OR 3.15, 95% CI 0.75–13.17). Similarly, the OR for shunt dependency in the monitor-first group was also lower (0.20) than the drain-first group (OR 4.94, 95% CI 0.54–45.58), but a larger confidence interval was noted (95% CI 0.02–1.87).


While most clinicians agree that an EVD should be placed in a patient with decreased level of consciousness when the presenting diagnosis is SAH, a standardized method of managing CSF drainage is absent. This study focused attention on describing the effect of two methods of management, both hypothesized to reduce the risk of cerebral artery vasospasm.

The hypothesized advantages of external CSF drainage after SAH are ICP control, and continuous removal of CSF and blood within the CSF that contains spasmogenic substances [2, 1114]. An early study (1991) found that continuous CSF drainage with the removal of a large amount of CSF was associated with increased vasospasm and hydrocephalus [10]. A more recent study found symptomatic vasospasm occurred in 17% of patients with lumbar CSF drainage and 51% of patients without lumbar drainage [13]. Our study did not demonstrate a difference in the incidence of vasospasm based on management strategy (drain-first group 66.7%; monitor-first group 53.9%). In addition to small sample size, a possible explanation for the lack of difference in vasospasm between the two groups is that the mean volume of CSF drainage per day did not differ between the two groups (drain-first 134 ± 17.1 mls; monitor-first 135 ± 22.8 mls, p = 0.79). If the volume of spasmogen-containing CSF cleared per day influences the likelihood of developing vasospasm then it is not surprising that did not see a difference between groups. As we do not have data on the prevalence of clinical vasospasm, we are unable to make any conclusions on the effects of the two treatment algorithms on microvascular vasospasm which may result in clinical changes without large vessel abnormalities as detected by TDC, CTA or catheter angiogram [15, 16].

Although this was a small, non-randomized observational pilot study, the groups were balanced in respect to traditional risk factors [4, 17]. Some known predisposing factors such as frequency of CSF sampling, site leaks, catheter irrigation, etc. were not analyzed in this study [1820]. Patients in the drain-first group experienced higher shunt rates and more complications related to the EVD although the differences were not statistically significant. Intermittent CSF drainage may prevent the formation of fibrous scars in the subarachnoid space by forcing the CSF into its normal pathways. This can clear the subarachnoid channels within blood-packed spaces and thereby allow persistent, sustainable CSF flow [4, 9, 19]. The incidence of EVD-related infections such as ventriculitis and meningitis have been reported between 0 and 27% [18]. In our study, about 13% of the drain-first patients developed ventriculitis versus 0% of those in the monitor-first group (p = 0.184). Duration of ventriculostomy and high ICP are known to be risk factors of EVD-related infections but were not different between these two groups.

The primary limitations of this study are sample size, lack of randomization and the internal validity threats of selection bias and instrumentation. The lack of randomization when enrolling subjects into the study significantly increased the risk for selection bias and the attainment of a non-representative sample. Selection effect was a major threat to external validity of this study because the subjects were assigned to continuous versus intermittent CSF drainage according to the preference of the attending neurosurgeon. Nurses and physicians who participated in the study cared for patients under both drain-first and monitor-first condition; as such, a recognized limitation is the failure to fully support an assumption of independence between the two groups. The limitations prohibit the ability to make strong inferences to other populations. However, this is the first study exploring this association, and data from this study provide sufficient evidence to support a randomized clinical trial.


We demonstrated the incidence of cerebral vasospasm after SAH was not different between two methods of ICP monitoring. However, shunt rate and ventricular catheter-related complications may be higher in the drain-first group than in the monitor-first group. This is a small sample size in an observational study and data from this pilot study provide for continued equipoise regarding a preferred method of ICP monitoring and CSF diversion in SAH. This supports further exploration of the relationship between CSF drainage techniques and the incidence of cerebral artery vasospasm in this population. A randomized large-scale study is indicated prior to drawing conclusions about which ICP monitoring technique is associated with prevention of cerebral vasospasm, reduced complications, or improved outcomes.


We are grateful to Nina Pluskowski.

Copyright information

© Springer Science+Business Media, LLC 2010