Cerebrospinal Fluid Drainage Trajectories Are Associated With Ventriculoperitoneal Shunt Placement After Aneurysmal Subarachnoid Hemorrhage

Abstract

BACKGROUND AND OBJECTIVES:

Predicting the need for ventriculoperitoneal shunt (VPS) placement after aneurysmal subarachnoid hemorrhage (aSAH) remains challenging, particularly among patients requiring prolonged external ventricular drainage (EVD). Existing predictors largely rely on static clinical or radiographic features and may not capture how cerebrospinal fluid (CSF) drainage evolves over time. We evaluated whether longitudinal CSF drainage trajectories, adjusted for EVD height, were associated with VPS placement after aSAH.

METHODS:

We performed a retrospective analysis of a prospective multicenter observational study of 149 adults with aSAH treated with EVDs at 6 US neurocritical care centers. Daily CSF drainage volume, EVD height, time from EVD placement, and VPS placement were analyzed using mixed-effects modeling to assess differences in CSF drainage trajectories over time.

RESULTS:

Thirty-four patients (22.8%) underwent VPS placement. Patients who underwent VPS placement had longer EVD duration and higher average daily and total CSF drainage during hospitalization. In longitudinal models adjusted for EVD height, CSF drainage trajectories diverged over time. Divergence emerged after day 8 from EVD placement, with CSF output declining more rapidly in patients who did not undergo VPS placement, whereas drainage remained more sustained in patients who underwent VPS placement. By day 18, height-adjusted CSF drainage was significantly higher among patients who underwent VPS placement. Among patients with EVD drainage for 18 days or longer, fewer than half underwent VPS placement, highlighting persistent uncertainty in this clinically challenging subgroup.

CONCLUSION:

Longitudinal CSF drainage trajectories, adjusted for EVD height, were associated with VPS placement after aSAH. Differences in CSF drainage trajectories emerged after day 8 and became more pronounced during prolonged EVD management. These findings suggest that longitudinal CSF drainage trajectories may complement static clinical and radiographic predictors in future shunt prediction models and studies of EVD weaning.

ABBREVIATIONS:

EVD

external ventricular drainage

mFS

modified Fisher score

VPS

ventriculoperitoneal shunt

WFNS

World Federation of Neurological Surgeons.

Aneurysmal subarachnoid hemorrhage (aSAH) remains a disease with significant morbidity and mortality despite advancements in specialized care and monitoring techniques within neurointensive care units. Among these contributors to morbidity and mortality is acute hydrocephalus, which may necessitate external ventricular drainage (EVD) placement for cerebrospinal fluid (CSF) diversion.^1,2 The reported incidence of acute hydrocephalus in the aSAH population ranges from 6% to 67%, with recent studies reporting approximately 20% of individuals developing it early in their course.^3,4 Methods to wean EVDs are highly variable and many patients ultimately require permanent CSF diversion through a ventriculoperitoneal shunt (VPS) system.⁵ Shunt dependency after aSAH is associated with high morbidity and mortality with both short- and long-term complications, such as acute obstruction or long-term infections. The avoidance of such a procedure with associated risk is of high importance.

Predictors of VPS dependency include acute hydrocephalus, presence of intraventricular hemorrhage, poor neurological grade on admission, older age, rupture of posterior circulation aneurysm or anterior communicating artery aneurysm, cerebral vasospasm, meningitis, surgical intervention approaches, rebleeding, or prolonged EVD placement, among others.^5-7 However, specific dynamic EVD features, such as days since EVD placement, and rate of CSF output over time adjusted for EVD height have not been well studied to ascertain their value and need in shunt prediction models and standardized weaning guidelines. This study aimed to assess whether dynamic trends in CSF output, adjusted for EVD height over time, predicted VPS dependency within the aSAH population. We also sought to identify a point in time at which CSF drainage features were most predictive of the need for a VPS.

METHODS

Study Population

This is a retrospective analysis of a prospective multicenter observational study at 6 neurocritical care centers in the United States focused on EVD wean strategies.⁷ The prospective study used rapid wean protocols, gradual wean protocols, or a mix of both. Institutional standard of practice was used to determine need for VPS through a multitude of factors such as persistent hydrocephalus, neurological examination, response to clamp trials, surveillance scanning, and CSF output. Since the study publication, additional patients have been enrolled prospectively and used in the analysis. This study has been approved by individual institutional review boards at each of the 6 participating centers. At each center, a waiver of informed consent was granted because the research posed minimal risk to participants, involved no procedures beyond routine clinical care, and could not practicably be carried out with individual consent. We used the Strengthening the Reporting of Observational Studies in Epidemiology checklist for study design and reporting of results.^8,9 The data of this study are available from the corresponding author upon request.

Study Sample

Adult patients age 18 years and older with suspected aSAH requiring EVD placement were included. Patients were excluded for the following reasons: nonaneurysmal etiology, transitioned to comfort measures only care or died in the neurointensive care unit before beginning an EVD wean trial, or EVD-associated epidural hematoma requiring craniotomy and EVD removal before any wean strategy (Figure 1). Participating centers placed EVDs primarily as a therapeutic intervention to manage acute hydrocephalus and were not placed solely to monitor intracranial pressure or correct cerebral perfusion pressure for vasospasm management. EVDs were not managed outside of the intensive care unit.

FIGURE 1.

Flow diagram for inclusion and exclusion of study population. A total of 200 patients admitted with subarachnoid hemorrhage were enrolled from 2017 to 2021. Adult patients age 18 years and older with suspected aneurysmal subarachnoid hemorrhage requiring external ventricular drainage placement were included. Of the 200 patients, 25 were excluded as they had a code status change to comfort care or died before any wean strategy, 24 were excluded as no culprit aneurysm was found, 1 was excluded for being a duplicate record, and 1 was excluded for having an external ventricular drainage–associated epidural hematoma requiring a craniotomy. Angio, angiogram; CMO, comfort measures only; VPS, ventriculoperitoneal shunt.

Data Collection

Patient demographics, including race and ethnicity, were collected through the electronic health record. Other recorded markers of injury severity included baseline severity via admission Hunt and Hess grade, World Federation of Neurological Surgeons (WFNS) grade, and the Glasgow Coma Scale score. Aneurysm data included treatment modality, location, and size. EVD data included height, duration, daily CSF drainage volume and appearance, and complications. We collected radiographic admission head computed tomography grades, which included the Fisher Score, Claassen Score, modified Fisher Score (mFS), and Graeb Score.

Outcomes

The outcomes assessed included VPS placement and daily CSF drainage volume, adjusted for EVD height, and trends over time. For patients who had more than 1 EVD placed during their hospital course, CSF drainage volume from the second EVD was excluded from the study.

Statistical Methods

Descriptive statistics were computed to describe the baseline characteristics of the patient sample overall and stratified by VPS placement. Comparisons between groups were performed using 2-sample t-tests for normally distributed continuous variables and Fisher's exact test for categorical variables with small sample sizes. Results are reported as mean and SD when appropriate. P-values <.05 were determined to be statistically significant. Missing data were handled with pairwise deletion.

A mixed-effects model was used for statistical analysis to determine the relationship between the continuous dependent variable CSF drainage output, and the fixed predictors of time (in days) from EVD placement (linear and quadratic, coded as continuous), time-varying linear EVD height (coded as continuous), a binary subject-level (time constant) predictor indicating whether each patient received a VPS or not (coded as nominal), and interaction terms for VPS and days (linear, quadratic). The random terms in the model were individual participants (coded as nominal) and their interaction with the linear term for time (coded as continuous). A residual repeated covariance structure and restricted maximum likelihood estimation was used.

A post hoc analysis was also conducted including the same independent variables within the mixed-effects model but substituting the dependent variable of daily CSF drainage output with the change in rate of CSF drainage over different times (ie, the delta CSF drainage output). Multiple delta CSF drainage output volumes were analyzed, including the delta between the CSF drainage of a given day compared with 1, 2, 3, and 7 days before. These delta CSF drainage output volumes were also analyzed as dependent variables among a smaller group of independent variables within the mixed-effects model, including EVD height, days since EVD placement, and VPS placement. An estimated quadratic regression equation was solved to determine at what day CSF output was predicted to be equal in patients who received a VPS vs those who did not. Statistical analyses were conducted using JMP Pro 15.2 (JMP Statistical Discovery LLC).

RESULTS

There were 200 patients who met initial inclusion criteria and were enrolled in the study (Figure 1). Of these, 51 were excluded, including 25 who were placed on “comfort measures only” or died in the intensive care unit before initiation of a wean trial, 24 patients who had no culprit aneurysm found, 1 patient who was found to be a duplicate record, and 1 patient who required surgical management of an epidural hematoma caused by EVD insertion.

Of 149 patients included in the analysis, 34 received a VPS while 115 did not. Most patients were White (64.7% and 76.5%) and female (82.4% and 69.6%) in both groups, with mean age 56.9 ± 14.5 years in those who received a VPS, and 55.5 ± 12.9 years in those who did not receive a VPS (Table 1). Both within the larger study population, and within the subset of 40 patients who underwent prolonged EVD drainage for 18 days or longer (Table 2), there was not a significant difference in baseline demographics between patients who received VPS and those who did not. Of the 6 centers in our multicenter study, 2 centers used rapid wean protocols, 3 centers used gradual wean protocols, and 1 center used a mix of both. There were no significant differences across institutions in terms of proportion of patients who underwent EVD drainage for 18 days or longer (Table 3, 95% CI [−0.20, 0.13], P = .21). In addition, when EVD weaning strategies were added into our mixed-effects statistical model, it did not meet statistical significance in predicting daily CSF drainage volume. When pooling data from all institutions, 26.9% of patients in the study underwent EVD drainage for 18 days or longer.

TABLE 1.

Baseline Demographic Characteristics of Entire Study Population

Characteristic	No VPS	VPS	P value
Number (n)	115	34	NA
Age (mean ± SD)	55.5 ± 12.9	56.9 ± 14.5	.6142
Female, n (%)	80 (69.6)	28 (82.4)	.1903
Race, n (%)			.4358
White	88 (76.5)	22 (64.7)
Black	13 (11.3)	4 (11.8)
Asian American	3 (2.6)	2 (5.9)
Other	7 (6.1)	4 (11.8)
Unknown	4 (3.5)	2 (5.9)
Ethnicity, n (%)			.6473
Hispanic	20 (17.4)	8 (23.5)
Non-Hispanic	89 (77.4)	25 (73.5)
Unknown	6 (5.2)	1 (2.9)
Hunt and Hess score, n (%)			.0225
1	4 (3.5)	3 (8.8)
2	34 (29.6)	3 (8.8)
3	43 (37.4)	15 (44.1)
4	20 (17.4)	11 (32.4)
5	14 (12.2)	2 (5.9)
Modified Fisher score, n (%)			.0322
1	7 (6.1)	2 (5.9)
2	9 (7.8)	3 (8.8)
3	61(53.0)	9 (26.5)
4	38 (33.0)	20 (58.8)
World Federation of Neurosurgical Societies			.1955
1	34 (29.6)	5 (14.7)
2	31 (27.0)	10 (29.4)
3	5 (4.3)	1 (2.9)
4	26 (22.6)	14 (41.2)
5	19 (16.5)	4 (11.8)
Treatment modality, n (%)			1.0000
Clip	24 (20.9)	7 (20.6)
Coil	82 (71.3)	25 (73.5)
Other	9 (7.8)	2 (5.9)
Average aneurysm size (mm ± SD)	6.0 ± 3.0	6.6 ± 3.2	.3302
Aneurysm location, n (%)
ICA	11(9.6)	3 (8.8)	1.0000
MCA	19 (16.5)	2 (5.9)	.1623
ACA	13 (11.3)	1 (2.9)	.1914
Vertebral	0 (0)	1 (2.9)	.2282
Basilar	9 (7.8)	7 (20.6)	.0541
PCA	5 (4.3)	1 (2.9)	1.0000
PComm	21 (18.3)	9 (26.5)	.3323
PICA	4 (3.5)	4 (11.8)	.0801
AComm	30 (26.1)	8 (23.5)	.8266
Other	3 (2.6)	1 (2.9)	1.0000
Angio-negative	0 (0)	0 (0)	—
Discharge disposition, n (%)			.0647
Home	43 (37.4)	7 (20.6)
Inpatient rehabilitation	54 (47.0)	16 (47.1)
Skilled nursing facility	17 (14.8)	11 (32.4)
Other medical hospital	1 (0.9)	0 (0)

ACA, anterior cerebral artery; AComm, anterior communicating artery; ICA, internal carotid artery; MCA, middle cerebral artery; NA, non applicable; PCA, posterior cerebral artery; PComm, posterior communicating artery; PICA, posterior inferior cerebellar artery; VPS, ventriculoperitoneal shunt.

Values reported as total (percentage) unless otherwise stated.

TABLE 2.

Baseline Demographic Characteristics of Study Population With External Ventricular Drainages in Place for 18 or More Days

Characteristic	No VPS	VPS	P value
Number (n)	21	19	NA
Age, y (mean ± SD)	52.5 ± 10.3	58.0 ± 10.0	.0949
Female, n (%)	15 (71.4)	16 (84.2)	.4570
Race, n (%)			.4814
White	11 (52.4)	11 (57.9)
Black	3 (14.3)	3 (15.8)
Asian American	0 (0)	2 (10.5)
Other	6 (28.6)	2 (10.5)
Unknown	1 (4.8)	1 (5.3)
Ethnicity, n (%)			.3874
Hispanic	7 (33.3)	4 (21.1)
Non-Hispanic	13 (61.9)	15 (78.9)
Unknown	1 (4.8)	0 (0)
Hunt and Hess score, n (%)			.5092
1	0 (0)	2 (10.5)
2	3 (14.3)	1 (5.3)
3	11 (52.4)	10 (52.6)
4	4 (19.0)	5 (26.3)
5	3 (14.3)	1 (5.3)
Modified Fisher score, n (%)			.2229
1	2 (9.5)	0 (0)
2	1 (4.8)	2 (10.5)
3	10 (47.6)	5 (26.3)
4	8 (38.1)	12 (63.2)
World Federation of Neurosurgical Societies			.5720
1	1 (4.8)	3 (15.8)
2	9 (42.9)	5 (26.3)
3	0 (0)	1 (5.3)
4	8 (38.1)	7 (36.8)
5	3 (14.3)	3 (15.8)
Treatment modality, n (%)			.5689
Clip	3 (14.3)	5 (26.3)
Coil	16 (76.2)	13 (68.4)
Other	2 (9.5)	1 (5.3)
Average aneurysm size (mm ± SD)	6.1 ± 3.1	6.8 ± 3.6	.5275
Aneurysm location, n (%)
ICA	1 (4.8)	2 (10.5)	.5962
MCA	3 (14.3)	1 (5.3)	.6071
ACA	3 (14.3)	1 (5.3)	.6071
Vertebral	0 (0)	1 (5.3)	.4750
Basilar	1 (4.8)	4 (21.1)	.1723
PCA	0 (0)	1 (5.3)	.4750
PComm	3 (14.3)	5 (26.3)	.4420
PICA	1 (4.8)	3 (15.8)	.3306
AComm	8 (38.1)	2 (10.5)	.0692
Other	1 (4.8)	0 (0)	1.0000
Angio-negative	0 (0)	0 (0)	—
Discharge disposition, n (%)			.6361
Home	4 (19.0)	4 (21.1)
Inpatient rehabilitation	9 (42.9)	11 (57.9)
Skilled nursing facility	7 (33.3)	4 (21.1)
Other medical hospital	1 (4.8)	0 (0)

ACA, anterior cerebral artery; AComm, anterior communicating artery; ICA, internal carotid artery; MCA, middle cerebral artery; NA, non applicable; PCA, posterior cerebral artery; PComm, posterior communicating artery; PICA, posterior inferior cerebellar artery; VPS, ventriculoperitoneal shunt.

Values reported as total (percentage) unless otherwise stated.

TABLE 3.

Distribution of Patients Across Each Institution Who Underwent EVD Drainage for 18 Days or Longer (Long EVD) and Eventually Required a VP Shunt

Institution	1		2		3		4		5		6		P-value
VP shunt	No	Yes	No	Yes	No	Yes	No	Yes	No	Yes	No	Yes
All patients	24	6	42	6	20	5	11	10	12	5	6	2	.0562
Long EVD patients	8	3	8	1	2	5	1	7	1	1	1	2	.0080
% Long EVD	33.3	50.0	19.0	16.7	10.0	100.0	9.1	70.0	8.3	20.0	16.7	100.0
	36.7		18.8		28.0		38.1		11.8		37.5		.2080

EVD, external ventricular drainage; VP, ventriculoperitoneal.

Patients with more severe clinical grades and those with higher degree of hemorrhage had higher rates of VPS placement. Those who received a VPS had a higher presenting Hunt and Hess Score (3 or 4: 76.5%), while those who did not receive a VPS presented with a lower Hunt and Hess Score (2 or 3; 67.0%; P = .03). In addition, those who received a VPS had a higher presenting mFS, with 58.8% presenting with a mFS of 4, compared with 33.0% in the non-VPS group (P = .03).

Finally, there were 8 cases of ventriculitis, which is consistent with what was found in the original cohort published in 2021. There were no additional cases of ventriculitis identified in the patients enrolled after the paper from the initial cohort was published. Eleven patients required EVD replacements due to clogged or non-patent EVDs. Of those 11 patients, clots were found in 6 patients and EVDs became nonpatent for unspecified reasons in 5 patients. Three patients who had EVDs in longer than 14 days required EVD exchanges. One of these was due to a clogged EVD mentioned above, the second was due to ventriculitis, and the third was due to CSF leak and tract hemorrhage.

EVD CSF Output in VPS vs Non-VPS Groups

The mean number of EVD days was significantly longer in those who received a VPS compared with those who did not (24.0 ± 9.7 days (95% CI [23.3, 24.8]) vs 16.4 ± 6.2 days (95% CI [16.1, 16.7]); P < .0001). Those who received a VPS had an increased mean CSF daily output (180.1 ± 115.3 (95% CI [171.0, 189.2])) compared with those who did not (164 ± 110 (95% CI [158, 170]); P = .003) and those who received VPS also had an increased total mean CSF output during their hospitalization (3258 ± 1536, 95% CI [2721, 3793]) compared with those who did not (2148 ± 1341, 95% CI [1900, 2395]; P = .0004) (Figure 2).

FIGURE 2.

Average CSF output during hospitalization in VPS vs non-VPS groups. Average A, daily and B, total CSF output (mL) was significantly higher in patients who received a VPS compared with those who did not. CSF, cerebrospinal fluid; VPS, ventriculoperitoneal shunt.

Forty of the 149 patients included in the analysis had 18 or more days of drainage with the initial EVDs placed during their hospitalizations. Twenty-one (52.5%) of these 40 patients did not end up receiving a VPS, while 19 (47.5%) did. In this subset of 40 patients, those who received a VPS had mean CSF daily output of 176 ± 117 (95% CI [165, 187]) compared with those who did not have a VPS (174 ± 108, 95% CI [163, 184], P = .73). In this same subset, those who received a VPS had total mean CSF output of 3906 ± 1479 (95% CI [3193, 4618]) compared with those who did not receive a VPS (3579 ± 1322, 95% CI [2977, 4180], P = .47).

Results from the mixed-effects model showed a significant interaction (95% CI [−0.86, −0.52], P < .0001) between the quadratic term for total EVD days and VPS placement, reflecting an increasingly rapid decline in daily CSF drainage output, more pronounced in the non-VPS group, even after adjusting for EVD height (Figure 3). As expected, increased EVD height was linearly associated with decreased CSF output (P < .0001, coefficient = −5.7, 95% CI [−6.1, −5.2]). Furthermore, observing this trend, increased CSF drainage correlated significantly with higher rates of VPS placement in the later stages of EVD treatment. Using a longitudinal approach, we found that after day 8, the patients who did not receive a VPS had progressively lower daily CSF drainage volumes compared with those who received a VPS. By day 18, patients with increased CSF drainage volume were more likely to require VPS, with a daily CSF output of 84 mL for those who received a VPS (95% CI [40, 127]) compared with 11 mL for those who did not receive a VPS (95% CI [−16, 39]; P < .05).

FIGURE 3.

Trend of CSF drainage volume over time (model predicted values). A, There is a steep drop in CSF output over time in those who did not receive a VPS. B, There is a more sustained production of CSF output when a VPS was required. CSF, cerebrospinal fluid; EVD, external ventricular drainage; VPS, ventriculoperitoneal shunt.

The discrepancy in CSF drainage between VPS and non-VPS groups became increasingly pronounced beyond day 18, contrasting with similar outputs before day 8 post-EVD placement (Figure 4).

FIGURE 4.

CSF output over time post-EVD placement in VPS vs non-VPS groups. A more rapid deceleration in CSF output across time was seen for patients who did not receive a VPS, compared with those who did. This was also seen after adjusting for EVD height as seen in the figure, illustrated for an EVD height at A, 10 cm H₂O and B, 20 cm H₂O. Curved lines represent model predicted values, with upper and lower shaded bands corresponding to 95% CIs. Approximately at day 8 post-EVD placement, there is similar CSF output between both groups, after which they diverge, and the trend is notable. By day 18, there was no longer overlap between the 95% CIs for the curves for the respective groups. CSF, cerebrospinal fluid; EVD, external ventricular drainage; VPS, ventriculoperitoneal shunt.

Radiographic Vasospasm

Radiographic vasospasm did not emerge as a statistically significant predictor of daily CSF drainage volume in the mixed-effects model (P = .47) or linear model (P = .41). In the subgroup of patients who had an EVD for 18 days or longer, the same trend was also observed in both the mixed-effects model (P = .88) and linear model (P = .29).

DISCUSSION

Our study suggests that patients who received a VPS had a significantly longer duration of EVD usage and higher daily and total CSF drainage throughout their hospital stay. However, notably, a dynamic CSF output trend emerged when comparing the daily output trends in those who received a VPS with those who did not. Longitudinal daily CSF drainage trajectories, adjusted for EVD height, diverged between these groups as early as day 8 post-EVD placement, and became more pronounced by day 18. These findings suggest that monitoring CSF output trends is an important factor in determining shunt dependency, in particular, in those with prolonged EVD needs. With nearly a third of the patients in our study (26.8%) requiring prolonged EVDs (≥18 days), our data also highlight that there is an unmet need for guidance on shunting for patients who undergo prolonged EVD courses. This dynamic predictor is of increased utility in the subset of population that required prolonged EVD placement as less than half in our study ultimately did not require a VPS.

For the first time, our results allow us to consider dynamic predictors beyond admission factors and isolated episodes of decompensation to include in prediction models to assist in EVD weaning strategies. Previous literature has largely focused on variables at presentation to predict long-term VPS needs, which may neglect the impact of changes the patient incurs due to either interventions or natural history of disease. Historically, predictors have included presence of intraventricular hemorrhage, poor neurological grade on admission, increased age, and rupture location of aneurysm, including posterior circulation or anterior communicating artery aneurysm.⁵ Beyond these admission factors, to account for variability as time progresses, decompensation or complications during hospitalization either through cerebral vasospasm, meningitis, or rebleeding have also been studied as predictors for need of permanent CSF diversion.⁵

Our study indicates that the change in height-adjusted daily CSF output can start to predict shunt dependency between day 8 and day 18, particularly in those with daily average outputs of greater than 180 mL/day. Beyond day 8 post-EVD placement is clinically impactful as it defines a point in time when CSF output trend has the most predictive value for shunt placement to help guide future standardized guidelines or protocols for EVD weaning strategies. Thus, future studies and predictive models should include these new dynamic features when evaluating for shunt dependency. Before this study, correcting CSF output trends for EVD height had not been well studied. As EVD height affects output, this is an important consideration that has shown impact and should be included in future study.

With a lack of high-level evidence and few standardized guidelines for EVD weaning strategies, current EVD discontinuation policies vary widely between institutions.¹⁰ This is evident in some centers reporting limited awareness of EVD weaning protocols.¹ Some favor a more gradual EVD wean approach,² while others have shown benefit with rapid EVD weaning.^11,12 However, the overall practice remains highly variable.^13,14 A survey conducted in 14 neurosurgical departments in Scandinavia reported general agreement toward attempting EVD discontinuation after 4 to 7 days from ictus when the patient is clinically stable, draining <150 mL/day, and intracranial pressure is continuously <15 mm Hg,¹ while an additional survey in the United States showed providers favoring a continuously open EVD with a gradual wean strategy.^15,16 This prospective multicenter observational study that used standardized weaning protocols showed that rapid wean was associated with fewer EVD days and fewer cases of nonfunctioning EVDs.⁷

The 2023 American Heart Association Guideline on aSAH reported inconclusive data on how to manage continuous vs noncontinuous CSF drainage through an EVD or lumbar drain for acute hydrocephalus and have highlighted the need for future research to address these practices.^5,17 In addition, the 2023 Neurocritical Care Society Guideline on aSAH concluded there is limited evidence to support rapid vs gradual weaning of the EVD, and further commented on the need for a multicenter randomized control trial comparing clearly defined protocols.¹⁸ Adding longitudinal predictors such as CSF flow output over time that capture initial severity along with the trajectory of the patient may be beneficial to consider as randomized control trials continue to be designed to address these questions.

Limitations

Our study has several limitations. First, since this is a retrospective post hoc analysis of a prospective data set, there is an inherent potential for bias. We recommend using the predictors and trends found in the study to guide design for future randomized control trials to further validate them. Second, our study showed association with initial disease severity using the Hunt and Hess grading scale consistent with the current literature; however, this correlation was not seen with the WFNS grade assessed on admission. This is likely attributed to the more detailed Hunt and Hess grading scale, which includes mild symptoms such as headache, rather than focusing on arousal and motor deficits, which are the hallmarks of the WFNS grade. Third, approximately 12.5% of enrolled patients were excluded from the analysis due to transition to comfort status or in-hospital mortality. This is a well-established challenge in conducting long-term outcome studies within the field of critical care.¹⁹ Fourth, additional analyses, such as area under the receiver operator curve, were unable to be conducted due to the use of continuous variable as the dependent factor in the mixed-effects model. Fifth, our results may have been affected by practice variations in EVD management between multiple centers, although both cohorts seemed to be well matched in terms of their weaning strategies. Finally, although we do see a new emerging trend not previously reported with adjusted CSF output for EVD height, we are still limited to a trend and not a daily numerical output value or cutoff to guide management. Further study is warranted to address these limitations.

CONCLUSION

Longitudinal CSF drainage trajectories, adjusted for EVD height, were associated with VPS placement after aSAH. Differences in CSF drainage trajectories emerged after day 8 and became predictive after day 18. This highlights the importance of trends beyond day 8 and provides helpful guidance regarding VPS placement in the subset of patients who have prolonged needs for EVD. Thus, longitudinal CSF drainage trajectories may complement static clinical and radiographic predictors in future shunt prediction models and studies of EVD weaning.

Funding

This work was funded by the National Institutes of Health (K08NS112601 and R01NS136224 to David Y. Chung and UL1TR001102 to Joseph J. Locascio). This work was conducted with support from Harvard Catalyst | The Harvard Clinical and Translational Science Center (National Center for Advancing Translational Sciences, National Institutes of Health Award UL1TR002541) and financial contributions from Harvard University and its affiliated academic healthcare centers. The content is solely the responsibility of the authors and does not necessarily represent the official views of Harvard Catalyst, Harvard University and its affiliated academic healthcare centers, or the National Institutes of Health.

Disclosures

Aman B. Patel received funding from Stryker. David Y. Chung received funding support from the American Heart Association and American Stroke Association (18POST34030369), the Andrew David Heitman Foundation, the Aneurysm and AVM Foundation, and the Brain Aneurysm Foundation. The other authors have no personal, financial, or institutional interest in any of the drugs, materials, or devices described in this article.

COMMENTS

In this multicenter study, the authors studying predictors of shunt dependency in patients with aSAH found that after day 8 of EVD, there was a notable decrease in CSF drainage in those not requiring VPS placement. Furthermore, by day 18, adjusted CSF drainage was higher in patients requiring VPS. CSF output trends beyond day 18 could be monitored to predict shunt dependency, as this study found many in this subset may not require a VPS. This study suggests that monitoring CSF drainage (volume, days, and trends) around day 8 of EVD can help predict VPS requirement. Although retrospective, the finding of a quantifiable finite day is noteworthy, and authors are commended for this analysis.

Raghuram Sampath

Clackamas, Oregon, USA