Introduction

Cerebral vasospasm is responsible for substantial morbidity and mortality in patients surviving initial treatment after subarachnoid hemorrhage [1]. Induced hypertension, hypervolemia and hemodilution (triple-H) therapy is intended to increase cerebral blood flow (CBF) to prevent or attenuate delayed ischemic neurological deficits in patients with vasospasm after SAH [2–5]. Despite widespread use in the United States, no large randomized controlled studies of triple-H therapy for the prevention or treatment of vasospasm have been conducted, and relatively few studies are available to guide clinical practice [6].

Recently, the physiological effects of individual components of triple-H therapy have been more carefully scrutinized [7, 8]. Induced hypertension has been demonstrated to consistently increase CBF and brain tissue oxygenation with a low incidence of complications [9, 10]. On the other hand, in a recent study, moderate hypervolemia was only rarely associated with increased brain tissue oxygenation, and complications were observed in as many as half of patients [11].

In a randomized trial comparing hypervolemic therapy with a strategy aimed at maintenance of euvolemia, Lennihan et al. [3] found that hypervolemic therapy increased cardiac filling pressures, but had no effect on blood volume, CBF or incidence of vasospasm when compared with euvolemic therapy. This study and others have also suggested that maintaining hypervolemia is difficult if not impossible, and that fluid balance and cardiac filling pressures are poor indicators of circulating blood volume [12]. Hypervolemia, if attained, may also result in potentially life-threatening complications that include hyponatremia, pulmonary edema, arrhythmia, congestive heart failure and cerebral edema [11, 13, 14].

Although induced hypertension seems to play a role in improving cerebral oxygenation, aggressive fluid administration may cause undue physiological stress with questionable benefits over maintaining euvolemia [2, 3]. In clinical practice, a wide variability exists in response to fluid administration; some patients maintain a euvolemic or negative fluid balance, whereas others tend to retain the fluids administered. We hypothesized that the prognosis of patients with SAH varies based on their early fluid balance, with differing susceptibilities to the complications of hypervolemia and/or vasospasm.

Methods

We retrospectively evaluated all adult patients admitted to a level-I trauma and stroke referral center in Seattle, WA with an aneurysmal SAH between May 1, 1999 and December 31, 2004. Data were collected from a hospital database of computerized medical and billing records. Transcranial Doppler (TCD) findings and head computed tomography (CT) imaging data were abstracted by chart review. The University of Washington Institutional Review Board approved the study and waived the need for informed consent.

All patients were diagnosed with an aneurysmal SAH documented by computed tomography (CT) and angiography. Patients with traumatic, venous or nonaneurysmal SAH, or pregnancy were excluded from the analysis. Two groups were delineated based on the early fluid balance: (i) the positive fluid balance group included patients whose cumulative fluid balance by ICU day 3 was positive and (ii) the negative fluid balance group included patients with zero or negative cumulative fluid balance by ICU day 3.

Management

Patients were admitted to the ICU and co-managed by neurosurgeons and neuro-intensivists based on established guidelines for the management of SAH [16]. In addition to aneurysm clipping or coiling within 72 h of hospital admission, all patients received oral or enteral nimodipine for 21 days. Patients were monitored for vasospasm with daily TCDs. The criteria for the interpretation of vasospasm were based on established guidelines [17, 18], as follows: (a) Middle cerebral artery (MCA) and internal carotid artery (ICA): (i) Normal: Time-averaged peak velocity (TAPV) <120 cm/s and hemispheric ratio <3; (ii) Mild MCA vasospasm: TAPV >120–149 cm/s and hemispheric ratio 3:5.9; (iii) Moderate MCA vasospasm: TAPV >150–199 cm/s and hemispheric ratio 3:5.9; (iv) Severe MCA vasospasm: TAPV >200 cm/s and hemispheric ratio >6; (b) Basilar artery: (i) Possible vasospasm (mild): TAPV >70 cm/s and hemispheric ratio >2; (ii) Moderate vasospasm: TAPV >85 cm/s and hemispheric ratio >2.5; (iii) Severe vasospasm: TAPV >85 cm/s and hemispheric ratio >3. Vasospasm of the anterior cerebral artery (ACA) was defined as a velocity of >130 cm/s, of the posterior cerebral artery (PCA) as a velocity of >110 cm/s and of the vertebral artery (VA) as a velocity of >80 cm/s. For the purpose of the study, patients were classified as having vasospasm if they had severe MCA and basilar artery vasospasm, and based on velocities criteria defined above in the ACA, PCA and VA.

All patients received at least 150 ml/h of total IV fluids, and fluid rates were increased to 200 ml/h if cerebral blood flow velocities on TCD were increased to values suggesting moderate-to-severe vasospasm. Patients with documented vasospasm had higher blood pressure and CVP targets corresponding to increasing severity of vasospasm (CVP goal of 8–10 mmHg for moderate vasospasm and CVP goal >10 mmHg for patients with severe vasospasm). Additional fluid boluses were given as needed to meet CVP and SBP targets. Blood products were used to keep the hematocrit between 24 and 30%. The systolic blood pressure (SBP) targets following clipping or coiling were 120–140 mmHg in patients with mild TCD vasospasm, 140–160 mmHg for moderate TCD vasospasm and >160 mmHg for severe TCD vasospasm. If patients were symptomatic, SBP was increased up to reversal of symptoms. In patients with high urine output requiring frequent fluid bolus and/or hyponatremia, fludrocortisone was used to increase renal sodium reabsorption and promote water retention. Vasopressors were initiated for patients not meeting their SBP goal.

Data Collection

ICU admission demographic, physiologic and laboratory values were retrieved from electronic medical and billing records. Troponin level peaks were dichotomized as >0.40 ng/ml or ≤0.40 ng/ml. If serum troponin was not measured, it was assumed not have peaked greater than 0.4 mg/ml. The Hunt-Hess, Fisher and Glasgow Coma Scale (GCS) were determined from the first head computed tomography (CT) and from ICU admission records. CT angiography and cerebral angiography records were used to determine the location of the aneurysm in the cerebral circulation. Admission head CT was also reviewed for the presence of preexisting nonacute stroke.

Net fluid balance was determined as the difference between total fluids-in and total fluids-out during the first 3 ICU days. Measured “fluids-in” included colloid, crystalloid solutions and blood products. Measured “fluids-out” included all urine output, intraoperative blood loss and NG tube output. Insensible losses were not accounted for. Both the change in weight and the change in serum creatinine between admission and hospital day 3 were recorded. The hematocrit, serum sodium, average MAP, intracranial pressure and SOFA score were recorded from the third hospital day. The peak troponin value was recorded in patients whose troponin rose to >0.40 ng/ml during the ICU stay. The use of CVP monitoring, mechanical ventilation, fludricortisone, vasopressors and blood products were recorded, and the average CVP and/or PaO2 to FiO2 ratios were recorded for patients who received CVP monitoring or were mechanically ventilated. Presence and location of the TCD vasospasm, requirement for angioplasty and occurrence of a new stroke at the 6-week follow-up head CT were noted. The average time to follow-up, the ICU and hospital lengths of stay, ICU discharge GCS and ICU and hospital mortality were recorded. The main outcomes were cerebral vasospasm demonstrated by TCD and a composite outcome of hospital mortality or new stroke on follow-up head CT.

Statistical Analysis

Bivariate associations between continuous variables and positive and negative fluid balance by ICU day 3 were investigated using two-sample t-test that allowed for unequal variances. Associations between categorical variables and the two fluid balance groups were investigated using the chi-square test statistics. Continuous variables are presented as mean ± standard deviation (SD); categorical variables are presented as frequency distributions.

Unadjusted regression models that included only the main predictor of interest (i.e., positive or negative fluid balance by ICU day 3) were fitted to estimate the association between fluid balance and outcome. The models were then adjusted for a priori-selected potential confounding factors including: age, Hunt-Hess score, Fisher grade, troponin elevation greater than 0.04 ng/ml at admission and mechanical ventilation at admission. The adjusted models included the 3-day fluid balance, significant predictors of mortality and the confounders.

Multivariable linear regression was also used to model length of ICU and hospital stay. Because length of stay was skewed, the data were log transformed before fitting the regression model. Estimates of length of stay comparisons are presented as ratio of medians with 95% confidence intervals.

To allow for the possibility that net fluid balance may change as length of stay increases, in sensitivity analysis, we explored associations between 7-day fluid balance and the outcome measures.

All P-values are two-sided with a 0.05 level of significance. A commercially available statistical program was used for all analyses (STATA for Macintosh, version 10.0; StataCorp, College Station, TX).

Results

Study Population

During the study period, 356 patients were admitted with a diagnosis of aneurysmal subarachnoid hemorrhage that satisfied the inclusion criteria detailed above. Of those, 135 patients had a cumulative fluid balance that was negative or zero, and 221 patients had a cumulative fluid balance that was positive at ICU day 3. Demographic characteristics of the study population are shown in Table 1. No important baseline differences were noted with respect to gender, age or admission weight, and the distribution of aneurysm location was similar between groups. Admission hematocrit, serum glucose, serum sodium and serum creatinine were also similar between the groups. Patients with positive versus negative fluid balance had similar MAP, and a similar proportion had troponin elevations greater than 0.40 ng/ml at admission. Patients with a positive fluid balance had significantly worse GCS and Hunt-Hess scores, but similar Fisher Grades.

Table 1 ICU admission characteristics of patients with subarachnoid hemorrhage stratified by 3-day cumulative fluid balance
Full size table

ICU Course

Physiological values on ICU stay day 3 and the cumulative fluid balance data by day 3 are shown in Table 2. Patients in the positive fluid balance group received more fluids than the negative fluid balance group, but cleared less fluids than the negative balance group by nearly 3 l (Table 2). The positive fluid balance patients had lower hematocrits, higher serum sodium values and higher SOFA scores on ICU day 3 when compared with patients with negative fluid balance. A higher proportion of positive fluid balance patients was mechanically ventilated and received blood products, whereas significantly fewer received fludrocortisone than patients with negative fluid balance. There were no differences in the serum creatinine and its change from baseline, the incidence of troponin peaks, the average MAP, intracranial pressure, use of CVP monitoring, PaO2/FiO2 ratios or use of vasopressors between the groups.

Table 2 Physiological values at day 3 of the intensive care unit stay stratified by cumulative fluid balance
Full size table

Study Endpoints

Vasospasm

Patient outcomes at hospital discharge and follow-up are shown in Table 3. TCD data were available in 126 out of 135 patients with negative fluid balance and 209 out of 221 patients with positive fluid balance. More patients with positive fluid balance by day 3 had documented TCD vasospasm, without differences in the location of vasospasm between the groups. Positive fluid balance on ICU day 3 was a predictor of TCD vasospasm both in unadjusted (OR 1.81, 95%CI: 1.16–2.82) and adjusted (OR 2.25, 95%CI: 1.37–3.71) analyses (Table 4). The frequency of angioplasty was no different between the two groups.

Table 3 Outcomes of patients with either positive or negative cumulative fluid balance by day 3
Full size table
Table 4 Unadjusted and adjusted odds ratios of TCD vasospasm, hospital death/new stroke and ratio of medians for hospital length of stay in patients with subarachnoid hemorrhagea
Full size table

Mortality and Stroke

ICU and hospital mortality were marginally higher in the positive fluid balance group but these differences were not statistically significant (P = 0.12 and P = 0.09, respectively). ICU discharge GCS was significantly worse in the positive fluid balance group, compared with the negative balance group, but without difference in the occurrence of a stroke during the hospitalization (Table 3).

About 11% of patients with positive fluid balance and 6% of patients with negative fluid balance had evidence of a new stroke on follow-up head CT (P = 0.10), with similar mean times to follow-up. A composite outcome of hospital death or new stroke was significantly higher in patients with positive fluid balance (P = 0.02). Prior to adjustment, cumulative fluid balance on ICU day 3 was significantly associated with a composite of hospital death or new stroke (OR 1.78, 95%CI: 1.11–2.87, Table 4). In adjusted analyses, 3-day positive fluid balance was not associated with higher odds of hospital death or new stroke relative to 3-day negative balance (OR 1.47, 95%CI: 0.85–2.54). Age, Hunt-Hess score and requirement for mechanical ventilation were all variably associated with increased odds of hospital death or new stroke (Table 4).

Length of Stay

The mean ICU and hospital length of stay were significantly longer in patients with positive fluid balance compared with patients with negative balance (Table 3). After adjustment, the ICU and hospital lengths of stay remained significantly different between the two groups (Table 4).

A sensitivity analysis comparing groups based on 7-day positive or negative cumulative fluid balance yielded similar findings as the 3-day fluid balance comparison.

Discussion

We observed important differences in patients with SAH based on their cumulative fluid balance early in the ICU stay. Patients with a positive fluid balance had a worse neurological presentation, a more complicated hospital course and longer ICU and hospital stays despite being managed with the same treatment protocol. A positive fluid balance was independently associated with the occurrence of TCD vasospasm and increased length of stay, but only marginally associated with the occurrence of new stroke or death.

All patients in the study received hypervolemic therapy following the same guidelines for fluid administration, yet we were able to separate two groups of patients with different responses. We observed that despite attempts to prevent natriuresis with fludrocortisone, some patients excreted the fluids administered and maintained a euvolemic or negative fluid balance. Additionally, patients with a positive fluid balance were prone to retain the fluids administered; they did receive cumulatively more fluids, but excreted cumulatively less, with similar renal function as the negative fluid balance group. These findings may indicate a difference in fluid handling between the groups that is determined by patient characteristics, and less so by the volume of fluid administered. A standardized protocol of fluid administration may not be optimal to fit all patients and particular attention should be paid in tailoring the amount of fluid administration based on the individual patient response.

Variability in fluid retention and elimination has been previously reported in patients following SAH [12], as indicated by the difficulty of maintaining a positive fluid balance or ever attaining true hypervolemia in some patients. Enhanced natriuresis is a common phenomenon in SAH patients, and could be one of the mechanisms responsible for a negative fluid balance despite fluid administration. Typically, these patients are thought to have either a cerebral salt wasting (CSW) physiology or a syndrome of inappropriate antidiuretic hormone secretion (SIADH). Although hyponatremia can be present in both of these scenarios, intravascular volume status may differ; typically, CSW is associated with hypovolemia and SIADH with relative euvolemia. Because CSW has been associated with vasospasm and worse outcome [15], there is a tendency to treat hyponatremic SAH patients with fluids more aggressively. One recent study, however, found that hyponatremic patients were more likely to be euvolemic with a SIADH rather than hypovolemic with CSW [16]. Conversely, CSW has also been commonly reported, and may be related to endocrine perturbations associated with increased sympathetic tone and the release of natriuretic peptides [17]. Our results are similar in that patients with negative fluid balance were more likely to be hyponatremic, but this was not associated with worse outcome. Because treatment was the same for both groups, it is possible that the negative fluid balance group fared better because the fluids administered were beneficial for restoring near-euvolemia, whereas the positive fluid balance group was already euvolemic and did not benefit from additional IV fluids.

Our results also indicate that neurological status at hospital admission may be a predictor of the response to early fluid administration. A positive fluid balance was associated with a significantly worse GCS, and higher frequency of poor Hunt-Hess grades compared with patients with a negative fluid balance. The physiologic mechanisms by which admission neurologic status are linked to subsequent fluid balance is not clear, but we noted significant differences in ICU course between the groups.

During the ICU stay, patients with a positive fluid balance had higher SOFA scores, a higher requirement for mechanical ventilation, and had lower hematocrit for which they received significantly more blood products. Whether the increased requirement for mechanical ventilation was directly related to fluid overload and pulmonary edema, or an association related to severity of illness is not clear from our data. Overall, the length of stay of patients with positive fluid balance was longer, and this difference persisted after adjustment for potential confounders.

The complications of fluid overload could be warranted if hypervolemic therapy increased CBF and reliably reduced the incidence of vasospasm/delayed ischemic events, but this has not been reliably demonstrated [3, 18–22]. Although the administration of IV fluids as a single bolus has been shown to increase CBF in some studies [18–20], induction of hypervolemia may not be associated with increased CBF, and could be accompanied by a higher incidence of complications [11]. When Lennihan et al. [3] compared euvolemic SAH patients to those randomized to receive hypervolemic therapy, there was no measurable difference in their blood volume, MAP or CBF. We did not measure CBF, but report similar findings with respect to CVP and MAP [3].

With respect to outcome, previous small randomized trials have found that compared with euvolemia, hypervolemic therapy does not improve vital or neurological outcome and does not decrease the incidence of TCD vasospasm [3, 23]. Our results are similar to those previously reported, with the exception of an increase in TCD vasospasm in the positive fluid balance group of our study. We are hesitant to expand on this observation as vasospasm was determined by TCD only, and although vasospasm appeared to be increased in the positive fluid balance group, there was no increase in the frequency of angioplasty. We also report a marginal association between positive fluid balance and poor outcome after adjusting for potential confounders. Despite adjustment, these findings may be a reflection of the worse neurological presentation of the positive fluid balance group at admission. It is also possible that the positive fluid balance group received more fluids because of early onset of TCD vasospasm. To limit this possibility, we chose to limit our cumulative fluid balance up to 72 h, because vasospasm is relatively uncommon before this time-point.

Other limitations of our study include the retrospective nature of the design. It is possible that the differences we observed in ICU course and outcomes were driven by differences in unmeasured confounding factors, though we attempted to adjust for these. The number of patients in the study group may also have been insufficient to identify differences in the measured variables.

Conclusions

The response to fluid administration is potentially predictable, depends on factors other than the quantity of fluid administered and is associated with worse clinical presentation at hospital admission. Taken together, these findings provide reason to question the utility of aggressive fluid administration for the management of aneurysmal SAH. The response to fluid administration should be integrated into the algorithm for continued fluid administered in such a way that limits uncorrected fluid therapy for patients with the worst clinical presentation. Clinicians can anticipate that patients with poor grade presentation might accumulate more fluids and need to be particularly vigilant to avoid fluid overload in this group. It is possible that greater hemodilution, longer mechanical ventilation and length of stay may be modifiable if fluid administration is tailored to patient response. Although fluid balance was not an independent predictor of mortality or new stroke, our results challenge the appropriateness of an approach using a preset amount of fluid administration, especially in SAH patients with poor presentation. Whether a tailored approach improves patient outcomes and reduces resource use remains to be determined.