Comparison of safety of lumbar drain versus external ventricular drain in symptomatic hydrocephalus following aneurysmal subarachnoid hemorrhage

Abstract

Background and objective

Cerebrospinal fluid (CSF) diversion is typically performed with an external ventricular drain (EVD) for symptomatic hydrocephalus (HCP) following subarachnoid hemorrhage (SAH). Lumbar drain (LD) has also been studied to reduce the incidence of vasospasm after SAH but not HCP. We performed a single center retrospective analysis to evaluate the safety of LD versus EVD for symptomatic HCP following aneurysmal SAH in a naturally randomized patient population.

Methods

Patients admitted for aneurysmal SAH who developed symptomatic HCP were treated with EVD or LD depending on neurosurgeon on call. Of the 10 neurosurgeons on call, five would place EVD in all patients while the other five would request LD be placed by interventional neuroradiology; however, the distribution on call was not evenly distributed. We retrospectively compared these two groups for drain complications and outcomes with Modified Rankin Scale (mRS).

Results

From 2018 to 2021, there were a total of 77 patients with aneurysmal SAH requiring CSF diversion for HCP. There were 56 cases of EVD placement and 21 cases of LD placement. Overall drain complications were 32.0% of cases with EVD and 9.5% with LD, p = .0773. EVD versus LD complications consisted of hemorrhage (1.8% vs 0%, p = 1.0000), infection (7.1 vs 0%, p = .5698), clogged (25% vs 0%, p = .008), dislodgement (1.8% vs 4.6%, p = .4737) and replacement (16% vs 4.8%, p = .2698). No case of cerebellar tonsillar herniation occurred. mRS between EVD versus LD obtained at baseline (0.3 vs 0.3, p = .3943), discharge (3.8 vs 2.7, p = .047), 90 days (2.9 vs 2.0, p = .060), and 1 year (2.6 vs 1.6, p = .081). One year mortality rates between EVD versus LD (26.8% vs 19.0%, p = .483).

Conclusion

Symptomatic HCP after aneurysmal SAH can be effectively and safely treated with LD. LD had lower overall complications than EVD with no hemorrhage, infection or malfunction. Further prospective randomized control study may be helpful in elucidating optimal CSF diversion for patients with symptomatic HCP.

Introduction

The estimated incidence of aneurysmal subarachnoid hemorrhage (SAH) worldwide is nine in 100,000 people per year. ¹ This disease process carries a significant risk of disability and death. The 30-day case fatality rate has been reported from 36% to 38%. ² Two serious consequences of SAH are hydrocephalus (HCP) and delayed cerebral ischemia (DCI). Systemic vasospasm is reported to occur in 20%–24% of aneurysmal SAH, a risk factor for DCI. ³ DCI is described to occur 6–9 days after SAH. ⁴ Screening and identification of vasospasm can be elusive in in cases when the patients are intubated and sedated. ⁵

HCP occurs in about 20–30% of aneurysm SAH but a range of 6–67% has been reported. HCP mostly develops acutely within 3 days but may also develop subacutely from 4 to 14 days. ⁶ Although the exact pathophysiology is unclear, HCP is thought to arise from either acute or chronic inflammation and with subsequent fibrosis of the arachnoid granulation, and maybe related to hemolysis of blood products.^6,7

The main treatment for HCP in the United States is external ventricular drain (EVD). Lumbar drain (LD) is increasingly adopted given more evidence that LD may also decrease DCI and vasospasm.^7,8

Permanent cerebrospinal fluid (CSF) diversion such as with ventriculoperitoneal shunt (VPS) is required in 17.4% of aneurysm SAH. Risk factors for VPS is age >60, intraventricular hemorrhage (IVH), high Hunt Hess score, rehemorrhage, and aneurysm location. ⁶

DCI represents an insidious sequela that can occur with or without cerebral vasospasm. ⁹ DCI can lead to severe neurologic deficits or death in 7–17% and the major source of morbidity and mortality after aneurysm treatment. ⁷ Since cerebral vasospasm maybe related to hemolysis of blood products, diversion of blood from cerebral arterial vasculature may reduce vasospasm.

LDs have been studied to reduce vasospasm but not as routinely used for treatment of HCP with aneurysm SAH in the United States. Most recently, Borkar et al. ⁷ found that LD decrease vasospasm and improve outcome in a randomized study in India but excluded patients treated for HCP and coiling. Safety concerns may limit LD for HCP and elevated intracranial pressure such as tonsillar herniation, over drainage, infection, clogging, and dislodgement.

At our medical center, neurosurgeons were divided between continuing to use EVD for all cases of HCP or open to use LD for those without IVH casting the ventricles. This created a natural randomization for temporary CSF diversion with either EVD or LD.

Materials and methods

This was a retrospective chart review from Kaiser Los Angeles Medical Center. The Kaiser Permanente Southern California Institutional Review Board (KPSC IRB) approval was obtained prior to the study. Medical records were queried for all patients who presented with SAH from 2018 through 2021. These records were further queried for presence of either an LD or external LD for CSF diversion. Our methods were carried out in accordance with the recommended guidelines and regulations as directed by the KPSC IRB, as it was a retrospective study, requirement of informed consent was waived by the KPSC IRB.

Patients from 2018 through 2021 who were admitted for aneurysmal SAH and determined to required CSF diversion were included. Patients with aneurysmal SAH were treated by the neurosurgeon on call. There were 10 neurosurgeons on call who decided on treatment of CSF diversion. Neurosurgeons on call varied in decision for CSF diversion ranging from only EVD to LD. LD was only considered in patients without intraventricular casting. Intraventricular casting was defined as severe IVH obstructing CSF flow. LD was placed under fluoroscopy by interventional neuroradiologists or by neurosurgery in OR (one case).

Subsequently, patients had cerebral angiogram and treated with aneurysm coiling or clipping. LDs were drained 10 ml per hour and then clamped since pressure measurements were not reliably measured with LDs

Modified Rankin Score were available at discharge and estimated from the chart for the other time periods: baseline, 90 days, and 1 year.

The modified Fisher grade was performed by single neuroradiologist (FT) blinded to treatment with EVD or LD.

Statistical analysis was performed using Kruskal–Wallis test for continuous variables and chi-square/Fisher's exact test for categorical variables. SAS Analytics software 9.4 was used.

Results

From 2018 to 2021, there were a total of 77 patients with aneurysmal SAH requiring CSF diversion for HCP. There were 56 cases of EVD placement and 21 cases of LD placement.

In terms of demographics, for EVD, Table 1, the average age was 58.8 versus 54.5 for LD, p = .485. There were 67.9% female in EVD group v 71.4% in LD group, p = .763. In terms of severity of presentation, there was statistically significantly higher Hunt Hess score in the EVD v LD; the Hunt Hess was 3.2 for EVD versus 2.5 for LD, p = .006. The modified Fisher grade were not statistically different with EVD v LD (3.9 v 4.0, p = .328).

Table 1.

Demographic and outcomes data of external ventricular drain (EVD) and lumbar drain (LD).

	EVD (N = 56)	LD (N = 21)	Total (N = 77)	p-value
Age				.4851 a
N	56	21	77
M (SD)	58.8 (12.63)	54.5 (16.64)	57.6 (13.85)
Median	60	60	60
Range	25.0, 81.0	21.0, 75.0	21.0, 81.0
mRS_baseline				.3943 a
N	56	21	77
M (SD)	0.3 (0.99)	0.3 (0.80)	0.3 (0.94)
Median	0	0	0
Range	0.0, 5.0	0.0, 3.0	0.0, 5.0
mRS_discharge				.0466 a
N	56	21	77
M (SD)	3.8 (1.95)	2.6 (2.29)	3.4 (2.11)
Median	4	3	4
Range	0.0, 6.0	0.0, 6.0	0.0, 6.0
mRS_90_days				.0598 a
N	56	21	77
M (SD)	2.9 (2.32)	1.8 (2.30)	2.6 (2.35)
Median	3	0	2
Range	0.0, 6.0	0.0, 6.0	0.0, 6.0
mRS_1_year				.0812 a
N	56	21	77
M (SD)	2.6 (2.45)	1.6 (2.40)	2.4 (2.46)
Median	2	0	2
Range	0.0, 6.0	0.0, 6.0	0.0, 6.0
Sex, n (%)				.7632 b
F	38 (67.9%)	15 (71.4%)	53 (68.8%)
M	18 (32.1%)	6 (28.6%)	24 (31.2%)
Hunt_Hess				.0060 a
N	56	21	77
M (SD)	3.2 (1.10)	2.5 (0.87)	3.0 (1.09)
Median	3	2	3
Range	1.0, 5.0	1.0, 4.0	1.0, 5.0
Modified Fischer				.328 b
N	56	21	77
M	3.88	3.95	3.91
Drain_problem, n (%)				.0773 b
0	38 (67.9%)	19 (90.5%)	57 (74.0%)
1	18 (32.1%)	2 (9.5%)	20 (26.0%)
Clogged, n (%)				.00843 b
N	42 (75.0%)	21 (100.0%)	63 (81.8%)
Y	14 (25.0%)	0 (0.0%)	14 (18.2%)
CSF_Infection, n (%)				.5698 b
N	52 (92.9%)	21 (100%)	73 (99.95%)
Y	4 (7.1%)	0 (0%)	6 (0.05%)
Hemorrhage, n (%)				1.0000 b
N	55 (98.2%)	21 (100.0%)	76 (98.7%)
Y	1 (1.8%)	0 (0.0%)	1 (1.3%)
Dislodged, n (%)				.4737 b
N	55 (98.2%)	20 (95.2%)	75 (97.4%)
Y	1 (1.8%)	1 (4.8%)	2 (2.6%)
Drain_replaced, n (%)				.2698 b
N	47 (83.9%)	20 (95.2%)	67 (87.0%)
Y	9 (16.1%)	1 (4.8%)	10 (13.0%)
VPS, n (%)				.7463 b
N	45 (80.4%)	18 (85.7%)	63 (81.8%)
Y	11 (19.6%)	3 (14.3%)	14 (18.2%)
Mortality, n (%)				.5654 b
N	41 (73.2%)	17 (81.0%)	58 (75.3%)
Y	15 (26.8%)	4 (19.0%)	19 (24.7%)

mRS: Modified Rankin Scale; VPS: ventriculoperitoneal shunt.

^aKruskal–Wallis p-value.

^bChi-square p-value.

Modified Rankin Scale (mRS) between EVD v LD were at baseline (0.3 v 0.3, p = .394), discharge (3.77 v 2.57, p = .047), 90 days (2.86 v 1.86, p = .060), and 1 year (2.63 v 1.62, p = .081). The mRS was statistically worse for EVD group at discharge with both groups improving and without statistical significance at 90 days and 1 year.

When comparing patients with better outcomes defined as mRS 0–2, there's a similar trend of proportionally higher better outcomes patients in the LD group. The percentage of mRS 0–2 between EVD v LD at discharge were (21.4% v 47.6%, p = .025), 90 days (44.6% v 66.6%, p = .068), and 1 year (51.8% v 76.2%, p = .111). There was no significant difference in the overall 1-year mortality rate between EVD v LD (26.8% v 19.0%, p = .565).

Overall drain complications were 32.1% with EVD and 9.5% with LD, p = .0773. EVD versus LD complications consisted of hemorrhage (1.8% v 0%, p = 1.000), infection (7.1% v 0%, p = .5698), clogged (25.0% v 0%, p = .0084), dislodgement (1.8% vs. 4.8%, p = .4737), and replacement (16.1% vs. 4.8%, p = .2698). All of the clogged EVD were resolved with flushing. Patients initially with EVD had replacement with EVD in 70% and LD in 30%. One patient initially with LD had replacement with EVD; in this case, 6 days after LD drain removed, patient developed recurrent HCP and EVD placed.

VPS was subsequently required in 19.6% of patients with EVD versus 14.3% in LD patients, p = .746. One patient with EVD had CSF leak after suboccipital craniectomy for posterior fossa decompression and subsequently required VPS. One case of replacement with LD after EVD was for recurrent HCP and CSF infection before VPS.

The one case of LD dislodgement occurred when patient was scratching his back 6 days after LD placement. The patient did not need another drain to be placed. No LD fracture was seen.

One patient with EVD developed HCP after removal, an LP was performed afterwards, and CSF culture grew Staphylococcus epidermis. The patient was started on antibiotics and an LD was subsequently placed for treatment of ventriculomegaly.

Discussion

Both EVD and LD effectively treated HCP resulting for aneurysmal SAH. There was an overall trend for lower complications with LD than EVD which was mainly driven by clogged EVD, however this was not statistically significant. All 14 of the clogged EVDs were resolved with flushing. This may be because only EVD was used for those cases where the ventricles were cast with hemorrhages which were not necessarily reflected in the Fisher grade. There were no hemorrhages or clogging with LD. One LD did get inadvertently dislodged; however, did not need to be replaced. There were overall four infections and none in LD group even though all patients were managed prophylactically with IV antibiotics. No subsequent CSF leak at the drain insertion site was seen in either group.

The study provided a natural randomization based on proclivity of the neurosurgeon on call for LD. There were over overall 2.6 times more EVD than LD reflecting the dominance of EVD as the treatment for symptomatic HCP with aneurysmal SAH. Half the neurosurgeons on call initially placed EVD in all their patients while the other half of neurosurgeons requested LD. Neurosurgeons who did not want to use LD were proportionally on call more during the study time period. However, this pattern changed over time and proportionally more LD were placed given the change in treatment philosophy among the neurosurgeons. A subsequent study may provide changes in EVD and LD treatment patterns and any change in complications or outcome.

LD is limited to communicating HCP. Although LDs were safely placed in patients with IVH, LD would unlikely be effective for large ventricular hemorrhages especially when casted. The EVD group had worse Hunt Hess with cases of severe ventricular hemorrhage casting which probably accounts for the higher rate of EVD clogging and later require VPS. Although modified Fisher grade were not statistically similar, this grading does not account for when ventricles are casted with hemorrhage.

Although widely feared, tonsillar herniation with lumbar puncture or drainage is extremely rare and likely only reported in cases with preexisting low lying cerebellar tonsils of at least 4 mm.^10–13 Paruchuri et al. ¹³ in their 55 patient intracranial hypertension series had no cerebellar tonsillar herniation after lumbar puncture. ¹⁰ There are two single case reports, Hoffman et al. ¹² and Borire et al., ¹³ of cerebellar tonsillar herniation after lumbar puncture for intracranial hypertension, but both had preexisting low lying cerebellar tonsils of 4 mm. Sullivan reported two cases of tonsillar herniation after lumbar puncture had neck pain with movement and focal neurologic deficits and therefore may have had preexisting tonsillar herniation. ¹⁴ There is only a single case report of cerebellar tonsillar herniation after LD, Sugrue et al., ¹¹ but that patient also had preexisting low lying cerebellar tonsils of 4 mm. However, nonfatal cerebellar tonsillar herniation has been reported with CSF leak after LD. ¹⁵ Therefore, these patients should be monitored for CSF leak.

Given the insertion of the LD in the lower back, LD may tend to be more easily dislodged although still relatively uncommon. Alqaim et al. ¹⁶ found similar 4% rate of LD dislodgement or fracture in their 100 patients series for endovascular aortic aneurysm repair, although they had a 16% rate of malfunction and no infectious complications. ¹⁶ While Estrera et al. ¹⁷ in their study of 1107 patient with LD for thoracic aortic surgery had 0.1% rate of fracture and 0.2% rate of meningitis. ¹⁷ Although Plotkin et al. ¹⁸ have been major complications (4.2%) of spinal hematoma with paraplegia (0.3%), intracranial hemorrhage (0.6%) and retained catheter tip (0.3%) reported on larger (309) LD case series for endovascular aortic repair, there were no major complications with LD seen in our study. None of these cases of LD for aortic aneurysm repair report cerebellar tonsillar herniation.

However, LDs for aneurysmal SAH may have higher rates of infections because of longer duration. Borkar et al. ⁷ found 16.7% rate of meningitis with LD; however, the group without LD also had meningitis rate of 6.7%.

Another limitation of LD is the difficulty to consistently measure CSF pressures and therefore removal of CSF based on pressure settings, while for EVD CSF removal was set to an intracranial pressure level, like 15 cm H₂O. Since pressure measurements were generally not effective with LDs, a set CSF volume was removed per time, like 10 ml per hour. However, this method may result in over and under CSF diversion and CT head scans were used to monitor HCP.

mRS was generally performed at discharge for the hospital stroke program and therefore obtained from chart review. However, since mRS is not generally available at other time points, mRS was estimated for baseline, 90 days, and at 1 year. Fortunately, most patients had long term relationships with the institution healthcare system, and therefore, baseline and follow up information was available.

Because other studies have studied LD for vasospasm, we did not directly study vasospasm in our study. These other studies specifically excluded patients with elevated intracranial pressure or HCP. ⁷ These studies demonstrated reduction in clinical vasospasm, vasospasm-related cerebral infarction, and hospital stay.^7,8 Similar to our study, in terms of outcome with mRS, Al-Tamimi et al. ⁸ found statistically lower mRS 4–6 with LD after aneurysmal SAH at day 10 but no difference than control at 6 months. ⁸ This suggests that patients can recover more fully in the long term regardless of treatment.

Conclusion

Symptomatic HCP after aneurysmal SAH can be effectively and safely treated with LD. Overall, LD tended to have lower overall complications than EVD but probably reflects more severe IVH in EVD group. Therefore, a prospective randomized control study may be helpful in elucidating optimal CSF diversion for patients with symptomatic HCP.