Effectiveness of Lumbar Cerebrospinal Fluid Drain Among Patients With Aneurysmal Subarachnoid Hemorrhage

Key Points Question Does prophylactic lumbar cerebrospinal fluid drainage improve clinical outcomes measured by the modified Rankin Scale score among patients with aneurysmal subarachnoid hemorrhage? Findings In this pragmatic randomized clinical trial including 287 patients at 19 sites in 3 countries, the rate unfavorable neurologic outcome was 32.6% in the lumbar drainage group (47 of 144) and 44.8% in the standard of care group (64 of 143), a significant difference. Meaning In this trial, among patients with aneurysmal subarachnoid hemorrhage, lumbar drainage improved clinical neurological outcomes at 6 months.


Subject Inclusion Criteria
 Aneurysmal SAH of all clinical grades  First aneurysmal SAH  Age: 18 years or older  Pre-morbid modified Rankin Scale score 0 ("no symptoms at all") or 1 ("no significant disability despite symptoms")  Aneurysm treatment performed during the first 48 hours after the initial hemorrhage.  Informed consent by the patient or his/her legal representative. In case neither the patient being capable of giving informed consent nor a legal representative is available, informed consent can be given by an independent physician neither involved in the patient's treatment nor in conducting the trial.

Subject Exclusion Criteria
 Subarachnoid hemorrhage of other than aneurysmal origin  No hemorrhage visible on initial CT scan (Fisher Grade I / modified Fisher Grade 0)  Pregnancy  Concurrent participation in another interventional trial (participation in an observational trial is not an exclusion criterion)  Life expectancy less than 1 year for other reasons than the current SAH  Other concomitant severe disease that would confound with treatment  Other clear contraindication for treatment with a lumbar drain (e.g. absent or compressed basal cisterns on the admission CT)
No information is available for the number of patients screened at the 14 remaining sites, which contributed 88 recruitments.

eFigure 2. Time Flow of Randomizations
Centers are ordered according to the date of their initiation for the EARLYDRAIN study. Numbers in parenthesis correspond to the number of randomizations contributed.

Invalid Recruitment
307 randomizations were performed for the EARLYDRAIN trial. In 20 instances, patients failed allocation and treatment according to their assignment.
Main reason for an allocation failure was lack or withdrawal of informed consent (four patients in each group). Two patients in the Lumbar Drain group and one in the Standard of Care group unexpectedly required antiplatelet treatment and therapeutic anticoagulation during or after aneurysm coiling. The eventual placement of a lumbar drain after the intervention was not feasible. Two patients in the Lumbar Drain group accidently got randomized twice due to a delay in response from the randomization server. Two patients in the Lumbar Drain group had aneurysm treatment later than 48 hours after the index subarachnoid hemorrhage, thus prohibiting starting lumbar drainage in due time as specified by the protocol. No source data was traceable for one patient in each group. In one patient in the Standard of Care group, no aneurysm was detected as bleeding source, and the patient was excluded from participating in EARLYDRAIN. In one instance, the patient was erroneously randomized and allocated to the Standard of Care group. In all these instances, no data was provided from the local investigators.
One patient randomized to the Lumbar Drain group was unable to receive the assigned intervention. Although initial not planned, double-platelet inhibition was required during aneurysm treatment with coiling. Therefore, placement of a lumbar drain was not feasible afterwards and no attempt was considered. This patient died on day eight after subarachnoid hemorrhage due to vasospastic infarctions. The data was excluded from all following analysis.
One patient randomized to the Standard of Care group was unable to receive scheduled aneurysm treatment due to rapidly developing brain edema. The patient died the day after admission. The data was excluded from all following analysis. (3) ( (1) (10) (1) (2) (3) (

Univariate analysis for outcome predictors in intention-to-treat data
For aneurysmal subarachnoid hemorrhage patients, known risk factors for worse outcome are age, and clinical and radiological severity grades. We analyzed the clinical data available on admission for association with the primary endpoint, unfortunate outcome measured by a modified Rankin Score larger than 2.

Intention-to-treat subgroup analysis
The following comparisons were performed to visualize the distribution of functional outcomes measured by the modified Rankin Scale in the intention-to-treat population. All p values were derived from logistic regression for a dichotomized mRS score of 0-2 vs. 3-6.

mRS at 6 months in patients treated elsewhere than the two largest centers eAppendix 1. Crossover Patients
Two patients randomized to the Standard of Care group were treated with a lumbar drain directly after aneurysm occlusion. Reasons for cross-over to the Lumbar Drainage group were:  assignment error (n=1)  request of neurological consultant (n=1) These two patients were not considered in the per-protocol analysis. Their data was included in the as-treated analysis in the Lumbar Drain group.
Further seven patients of the Standard of Care group received a lumbar drain for clinical reasons at day 3 or later during the first eight days. Their data was kept in the per-protocol and as-treated analysis as Standard of Care.
35 patients allocated to the use of a lumbar drain after aneurysm occlusion were not treated accordingly. This data was excluded from the per-protocol analysis. It is available in the as-treated analysis in the Standard of Care group.
In 13 of these 35 patients, no attempt to place a lumbar drain was performed. Reasons provided were:  technically not feasible (n=7)  assignment error (n=3)  risk assessment by clinical judgment (n=3) In 22 patients, a lumbar drain was inserted, but no relevant drainage was performed. Reasons were:  risk due to ICP difference between EVD and LD (n=1)  risk assessment by clinical judgment (n=5)  clotting of drainage (n=8)  no reason given (n=8) A comparison of cross-over patients with the regularly treated cohort is given in eTable 10. Only one cross-over patient had a posterior circulation aneurysm. Otherwise, cross-over patients had similar demographic and hemorrhage severity characteristics to patients treated according to protocol. No pattern predisposing to crossover was noted.

eAppendix 2. Per-Protocol Analysis
In the Lumbar Drain group, 109 patients (75.7%) received a lumbar drain after aneurysm treatment and sufficient lumbar drainage as per protocol for the next eight days. In 22 of the Lumbar Drain group patients (15.2%), lumbar cerebrospinal fluid was drained less than specified; in 13 patients (9%) no lumbar drain was placed. Patients with less than specified or no drainage were excluded from the Lumbar Drain group in perprotocol analysis.
134 patients (93.7%) in the Standard of Care group were treated without lumbar drain in the first eight days. Seven patients (4.9%) got a lumbar drain during the first week of treatment but received less than 480 ml of drainage. These patients were kept in the Standard of Care group for per-protocol analysis. Two Standard of Care patients (1.4%) underwent placement of a lumbar drain after aneurysm treatment with consecutive drainage of more than 480 ml and were therefore excluded from per protocol analysis.

Univariate analysis for outcome predictors in per-protocol data
For aneurysmal subarachnoid hemorrhage patients, known risk factors for worse outcome are age, and clinical and radiological severity grades. We analyzed the clinical data available on admission for association with the primary endpoint, modified Rankin Score larger than 2.
Clinical grading according to Hunt and Hess or WFNS scales showed collinearity. The same was found for radiological gradings. Models with lower Akaike Information Criterion values were chosen for multivariate analysis. Accordingly, in multivariate outcome assessment, adjustment for baseline imbalances was performed with the parameters age, Hunt-Hess grade larger than 3 and intracerebral or intraventricular hemorrhages (

Univariate analysis for outcome predictors in as-treated data
For aneurysmal subarachnoid hemorrhage patients, known risk factors for worse outcome are age, and clinical and radiological severity grades. We analyzed the clinical data available on admission for association with the primary endpoint, unfortunate outcome measured by a modified Rankin Score larger than 2.
Similar to the per-protocol data, clinical grading according to Hunt and Hess or WFNS scales as well as radiological grading showed collinearity. Models with lower Akaike Information Criterion values were chosen for multivariate analysis. Accordingly, in multivariate outcome assessment, adjustment for baseline imbalances was performed using the parameters age, Hunt-Hess grade larger than 3 and intracerebral or intraventricular hemorrhages (Table 2, main manuscript). These parameters were also used in regression models for assessment of the risk of infection for external ventricular drains and lumbar drains.

eAppendix 4. Adverse Events
Vasospasm data is given in Table 2 in the main manuscript and eTables 3, 13, and 16 in this supplement.
The development of secondary infarctions other than being related to aneurysm occlusion was considered a key secondary endpoint in EARLYDRAIN. This data is available in Table 2 in the main manuscript and eTables 13 and 16. Further analysis on the relations of vasospasm, infarctions and modified Rankin scores at six months are in this appendix, eTables 4 to 9.
The requirement of a permanent cerebrospinal fluid shunt implantation for treatment of hydrocephalus was regarded an adverse event by some investigators and not separately reported by others. Numbers for permanent shunt necessity are given in Table 2 in the main manuscript and eTables 13 and 16 in this appendix.
One patient in each group was diagnosed with vitreous ocular hemorrhage, also known as Terson's syndrome.
One patient in the Lumbar Drain group was noted to have a local skin abscess at the lumbar punction site, requiring surgical excision.
One patient in the Lumbar Drain group had the lumbar catheter torn off when extraction was attempted, requiring surgery for removal of the remnant.
Patients with aneurysmal subarachnoid hemorrhage frequently experience a multitude of medical complications during their clinical course, including fever, pneumonia, electrolyte disturbances, cardiac failure, hypo-and hypertension. Per definition, these were not regarded as adverse events in EARLYDRAIN and investigators were encouraged to omit separate recording.
Data on the frequency infections is presented in Table 2 in the main manuscript and eTables 13 and 16. Analysis of risk factors for infections is given in eTables 20 and 21.

Assessment of infection risk
Clinical suspicion of infection included but was not restricted to central nervous system involvement including device-associated meningitis. We analyzed the clinical factors possibly associated with the development of infection. For multivariate assessment of the risk for the development of infections, we selected age, Hunt-Hess grade larger than 2 and intraventricular or parenchymal hemorrhage as predictors, according to the lowest values of the Akaike Information Criterion. Use of an external ventricular drain and use of a lumbar drain were forced into the multivariate statistical model, regardless of AIC in univariate analysis.