Pilot study of volume contracted state and hospital outcome after stroke

Abstract

Background

An increasing body of research suggests that acute stroke patients who are dehydrated may have worsened functional outcomes. We sought to explore the relationship between a volume contracted state (VCS) at the time of ischemic stroke and hospital outcomes as compared with euvolemic patients.

Methods

We enrolled a consecutive series of ischemic stroke patients from a single academic stroke center within 12 hours from stroke onset. VCS was defined via surrogate markers (blood urea nitrogen/creatinine ratio >15 and urine specific gravity >1.010). The primary outcome was change in NIH Stroke Scale (NIHSS) score from admission to discharge. Multivariable analyses included adjustment for demographics and infarct size.

Results

Over an 11-month study period, 168 patients were eligible for inclusion. Of the126 with complete laboratory and MRI data, 44% were in a VCS at the time of admission. Demographics were similar in both the VCS and euvolemic groups, as were baseline NIHSS scores (6.7 vs 7.3; p = 0.63) and infarct volumes (12 vs 16 mL; p = 0.48). However, 42% of patients in a VCS demonstrated early clinical worsening, compared with 17% of the euvolemic group (p = 0.02). A VCS remained a significant predictor of worsening NIHSS in adjusted models (odds ratio 4.34; 95% confidence interval 1.75–10.76).

Conclusions

Acute stroke patients in a VCS demonstrate worse short-term outcomes compared to euvolemic patients, independent of infarct size. Results suggest an opportunity to explore current hydration practices.

An estimated 40%–60% of ischemic stroke patients are dehydrated, volume contracted, or both at the time of hospital presentation.^1–4 Physicians may be hesitant to provide aggressive rehydration, citing congestive heart failure, intracranial pressure elevation, or hypertension as primary concerns. A growing body of evidence suggests a relationship between dehydration and poor outcome measured by risk for neurologic fluctuation, worsened hemispatial neglect, increased hospital complications, and hospital disposition after stroke.^3–10 though only one was conducted in the United States.⁶ There are even fewer data evaluating specific fluid regimens to treat these patients.^11,12 One issue in designing such a trial is that assessment of hydration status remains subjective.^13,14 This study addresses that gap by exploring the relationship between laboratory markers of hydration and stroke outcome in an urban comprehensive stroke center.

The gold standard for measurement of intravascular volume relies on invasive techniques (e.g., pulmonary artery catheters), which are impractical in the stroke population.¹⁵ Subjectivity in the bedside assessment of a patient's hydration status contributes to variability in practice and likely impedes research in this area. In the clinical arena, clinicians often rely on laboratory markers as indicators of hydration status.¹⁶ There is a need to further examine the relationship between these laboratory markers and clinical outcome in a stroke population, who are at high risk for volume overload and intracranial hypertension.

We sought to (1) define a volume contracted state (VCS) using surrogate laboratory markers; (2) compare stroke severity based on NIH Stroke Scale (NIHSS) score and infarct size on MRI based on volume status; (3) compare short-term clinical outcomes based on a patient's volume status at the time of stroke; and (4) explore disrupted cerebral perfusion as the mechanism behind these relationships. We hypothesized that a VCS at the time of acute ischemic stroke would contribute to a worse short-term clinical outcome, defined as change in NIHSS score between the time of acute stroke and hospital discharge.

Methods

Between July 2013 and June 2014, a consecutive series of acute ischemic stroke patients were enrolled into this observational cohort study if they presented within 12 hours from onset of symptoms to the Johns Hopkins Hospital and demonstrated acute ischemic stroke on MRI. For pathophysiologic reasons, patients with baseline history of renal failure, admission creatinine ≥2.0, or active infection at the time of presentation were excluded.

Standard protocol approvals, registrations, and patient consents

The study was approved by the Johns Hopkins Institutional Review Board and consent for participation was waived since there was no change from the standard of care.

Clinical variables

The majority of ischemic stroke patients at our institution have laboratory testing performed upon arrival to the hospital, allowing for evaluation of blood urea nitrogen (BUN) and creatinine; a subset of patients have urine specific gravity measured, based on individual physician practices. BUN/creatinine ratios were explored as a continuous variable. BUN/creatinine ratios were then dichotomized by VCS if the admission laboratory studies revealed BUN/creatinine ratio >15 and urine specific gravity >1.010 as in prior reports.^5,6,17 All patients underwent clinical assessments including baseline NIHSS and daily neurologic examinations for 4 days or until the time of discharge if sooner. Clinical examinations were converted to daily NIHSS scores as per previously validated methods.¹⁸ Nursing documentation and daily clinical notes were reviewed for report of any fluctuation in neurologic status, and NIHSS was again calculated for those changes in clinical condition. The primary stroke team was unaware of the study hypothesis, so care was indistinguishable across groups.

The primary outcome was change in NIHSS score, calculated as admission NIHSS score subtracted from discharge NIHSS score, with negative values representing improvement and positive values representing worsening. This change in NIHSS score was then categorized into quartiles, ranging from improvement in NIHSS score >4 points (quartile 1) to worsened NIHSS (quartile 4) at discharge as compared to on admission. Secondary outcomes included significant fluctuation in neurologic status (even transient), defined a priori as a change in NIHSS score >3 points at any point during the hospital stay.² For the subset of patients who had perfusion weighted imaging as part of their baseline MRI, secondary outcomes included evaluation for diffusion–perfusion mismatch.¹⁹ Covariates of age, sex, infarct size, and baseline glucose were planned a priori for multivariable regression modeling.

Neuroimaging analysis

In order to explore a potential mechanistic explanation to the relationship between hydration and functional outcomes, we used MRI to quantify stroke size and estimate cerebral perfusion. MRI volumetrics were calculated using Olea Sphere software by a single primary rater (M.N.B.) who was blinded to both diagnosis and hydration status. Both the lesion volume based on diffusion-weighted imaging (DWI) and volume of hypoperfusion based on perfusion-weighted imaging (PWI), for participants with PWI sequences, were calculated and compared for mismatch using an apparent diffusion coefficient threshold of 600 and time to peak of 6.0 seconds delay relative to the intact hemisphere.

Statistical analysis

Statistical analysis was conducted using Stata version 12.0 for Macintosh. For all analyses, the primary dependent variable was quartile of NIHSS change. In logistic regression analyses, presence in the worst quartile of NIHSS change from admission to discharge was evaluated as the outcome, with individuals in any of the first 3 quartiles representing the reference group. The odds ratios were adjusted for the following demographics and potential confounders: age, sex, DWI lesion volume, and baseline serum glucose, and are reported with 95% confidence intervals (CIs). To evaluate outcome using the full range of the 4 quartiles of NIHSS change, we also used ordinal logistic regression with quartile of NIHSS change as the outcome of interest. Quartile 1 represented patients who had an improved NIHSS score >4 points in the first 4 days; quartile 2, improved NIHSS score 1–3 points; quartile 3, no change in NIHSS score; and quartile 4, worsened NIHSS score between admission and discharge from hospital. We also evaluated the amount of change in the NIHSS, on a continuous scale, in linear regression models, adjusted for the same covariates. Finally, in some analyses, we considered best quartile of NIHSS change, representing improvement, as the outcome, with presence in the other quartiles as the reference group.

Binary secondary outcomes were analyzed with logistic regression and are reported as unadjusted and adjusted odds ratios with 95% CIs; Spearman correlations were used to evaluate correlations between baseline characteristics (NIHSS and lesion volume, comparing those with and without volume contraction). We calculated that a sample of 172 patients would yield a power of 80% at a significance level of 0.05 to test the proposed hypothesis and detect a difference between the volume contracted and euvolemic groups.

Results

During the 11-month enrollment period, 306 ischemic stroke patients were screened for enrollment. Of those, 168 patients met inclusion criteria, 126 (75%) of whom had full laboratory panels (both BUN/creatinine and urinalysis) and MRI.

Using the definition of VCS as defined above (using both serum and urine markers), 55/126 (44%) patients were in a VCS at the time of their admission to the hospital. The mean age of study participants was similar in the volume contracted and euvolemic groups (66 vs 63 years; p = 0.22). There were no statistically meaningful differences in demographics or relevant baseline characteristics including initial blood pressure or heart rate between the 2 groups (table 1). Similar amounts of IV fluids were administered over the 4-day observation period (median 1.25 L in the volume contracted group and 1.72 L in the euvolemic group; p = 0.64).

Table 1.

Demographics of the study population meeting inclusion criteria

VCS and stroke severity

Initial stroke severity was not different in the volume contracted vs euvolemic group (median NIHSS 6.7 vs 7.3, respectively; p = 0.63). Average infarct size on MRI was 12 mL in the volume contracted group and 16 mL in the euvolemic group (p = 0.48). There was no meaningful difference in the diffusion–perfusion mismatch between groups. The relationships between NIHSS score and lesion volume on baseline MRI did not significantly differ based on volume status (ρ = 0.39 [volume contracted] vs 0.50 [euvolemic]).

VCS and hospital outcome

Patients who were volume contracted were more frequently in the worst quartile of change in NIHSS (quartile 4) than were euvolemic patients (42% vs 17%; p = 0.02) (figure). This difference between volume contracted and euvolemic patients persisted with multivariable analysis, where the volume contracted group had a 4.34 times increased odds of being in quartile 4 for change in NIHSS score after adjustment for age, sex, glucose, lesion volume, and baseline NIHSS (95% CI 1.75–10.76). Results were similar when evaluated across all 4 quartiles of NIHSS change (table 2), with over 2-fold increased odds of being in a worse quartile of NIHSS change in volume contracted patients (compared to those who were euvolemic).

Figure. Change in NIH Stroke Scale (NIHSS) score between hospital admission and discharge.

Distribution of quartiles of change in NIHSS between hospital admission and discharge for volume contracted vs normal volume states. Percentages shown are out of each total volume status group (e.g., 24% of patients in volume contracted state were in quartile 1, suggesting most improved NIHSS).

Table 2.

Relationship between volume contraction status and amount of NIH Stroke Scale (NIHSS) change

On average, when continuous change in NIHSS was evaluated, the change in NIHSS for the volume contracted group was 1.62 points higher (adjusted; 95% CI 0.07–3.17, consistent with less improvement or with worsening) than was the NIHSS change in the euvolemic group. Rates of neurologic fluctuation did not vary between the 2 groups (25% volume contracted vs 34% euvolemic patients had fluctuation in NIHSS >3 points) (p = 0.335).

Discussion

These results from a single-center study suggest that a VCS at the time of ischemic stroke is associated with an increased odds of worse short-term outcome after acute ischemic stroke, even after adjustment for other confounders and markers of stroke severity (e.g., infarct size). This study expands on prior studies in that this series, conducted at a comprehensive stroke center, includes only MRI-confirmed ischemic stroke patients, reports PWI in the volume contracted and euvolemic groups, and measures amount of IV fluids administered. These findings support prior reports in the literature^3–10 and suggest a potential opportunity for improvement in our hydration strategy for patients who are volume contracted at the time of presentation with ischemic stroke. This study also validates results from several prior reports demonstrating that a large percentage of ischemic stroke patients are volume contracted at the time of their stroke.^{1–6,20–22}

Though yet unproven, there are several biological theories about why a VCS at the time of stroke could influence functional outcome. These include reduction of cerebral perfusion during critical periods of failing autoregulation, increased overall viscosity disrupting microcirculatory pathways, and the dilution and circulation of inflammatory cells that may accelerate cerebral edema in the region of infarction.^23-30 In an experimental animal stroke model, restricted food and water intake increased mortality rates to 59% compared with the 15% rates in animals with normal intake. In addition, their results demonstrated that animals do not drink spontaneously poststroke vs naive mice (p < 0.01).³¹ Treatment with IV fluids could reduce these physiologic effects and improve functional outcome, though no study has yet demonstrated the effect of saline on stroke outcome in humans.⁷

This study was novel in that we objectively measured volume status acutely, at the time of presentation (presumably, before the stroke itself could have affected hydration status), only included MRI confirmed stroke patients, evaluated the association between hydration and outcome independent of infarct size, and measured outcome prospectively in a comprehensive stroke center. By using infarct volume, we hoped to have addressed confounding by stroke size. Interestingly, larger infarct volume was independently associated with less worsening in the NIHSS; this may reflect a ceiling effect in that patients with larger strokes had higher NIHSS at onset, leaving them less room to worsen. Alternatively, this result could reflect the fact that the largest strokes were due to embolism, in which case completed infarction likely occurs earlier after stroke onset due to underdeveloped collateral circulation. Finally, though underpowered, these data did not suggest a statistically meaningful difference in diffusion to perfusion mismatch between volume contracted and euvolemic stroke groups. Diffusion/perfusion mismatch on MRI remains a clinically relevant variable that will continue to be measured moving forward as we continue to explore the mechanistic reasons behind the relationship of hydration status and functional outcome after stroke.

Several limitations are notable. First, the sample size with all laboratory tests and complete MRI data is relatively small; laboratory tests are standard of care in stroke admissions, but urinalysis is not performed in all admitted patients. Future studies with a larger sample size are planned. Next, the timing of MRI acquisition was not always immediate upon arrival to the hospital, and perfusion scans were not available in the entire group. There was a non–statistically significant, but perhaps clinically meaningful, imbalance in thrombolytic administration between the volume contracted and euvolemic groups. Future randomized trials will need to balance this variable between groups in order to understand hydration-related functional outcomes and include 3-month modified Rankin Scale scores to assess for clinically meaningful differences. Finally, we acknowledge that our markers of hydration and volume status are not directly measured and that other technologies to examine intravascular volume status (e.g., pulmonary artery catheterization that requires right heart catheterization) may serve as more of a direct measurement. Our study shows an association between commonly used laboratory markers of VCS and short-term clinical stroke outcomes. This finding is important as we attempt to find objective surrogate markers that are noninvasive, clinically relevant, and easily adaptable internationally for use in future studies to evaluate the relationship between volume status and long-term stroke outcomes.

Despite the limitations, our data support the hypothesis that a VCS at the time of stroke is associated with worse short-term outcome. This is important as early neurologic worsening after stroke has been repeatedly demonstrated to affect longer-term outcome, most recently in an analysis of 368 patients who demonstrated a 35 times increased odds of death or dependency if they demonstrated early neurologic worsening.³² Different from prior studies, we did not see a high rate of transient neurologic fluctuation in this cohort, as previously reported.² Thus, our study provides essential markers to allow us to address persistent questions specific to the relationship between fluid volume and outcomes after stroke. While prior stroke studies have explored volume expansion or hemodilution to test a variety of other hypotheses, no study has tested whether hydration vs placebo has any benefit after stroke. Such a study, using objective markers of hydration status, is needed in order to refine recommendations about dose, duration, and effect of such a low-cost, broadly available therapy.

A large percentage of ischemic stroke patients are volume contracted at the time of their presentation to the hospital, and these patients are less likely to improve in the early treatment period. Future studies will need to evaluate potential mechanisms for this association, and whether volume contracted status at the time of stroke carries any risk for longer-term worsened outcomes. Such a relationship could warrant investigation of more precise rehydration strategies for improved early recovery of this population of patients.

Author contributions

M.N. Bahouth: conceptualized the study, designed, collected data, interpreted data, and drafted and revised the manuscript. A. Gaddis: assisted with data collection and revised the manuscript. A. E. Hillis: assisted with study design and revised the manuscript. R.F. Gottesman: assisted with study design, interpreted and analyzed data, and revised the manuscript.

Study funding

This work was supported by NIH/NINDS R01NS047691 to A.E.H. and by R25 NS065729 to A.E.H. with supplement to M.N.B.

Disclosure

M.N. Bahouth is supported by a Richard Starr Ross Clinician Scientist Award from Johns Hopkins School of Medicine. A. Gaddis reports no disclosures. A.E. Hillis serves on a DSMB for Axovant; serves as an Associate Editor for Stroke, Aphasiology, and Practice Update Neurology; receives publishing royalties for Handbook of Adult Language Disorders (Taylor & Francis, 2015); serves as a consultant for The Charles Dana Foundation; and receives research support from NIH (NIDCD, NINDS). R.F. Gottesman serves on a DSMB for Genentech; has received funding for travel to attend a Cardiothoracic Surgery Symposium; serves as an Associate Editor for Neurology®; receives florbetapir isotope (¹⁸F-AV-45) for investigator-initiated research by Avid Radiopharmaceuticals, a wholly owned subsidiary of Eli Lilly; and receives research support from NIH/NIA. Full disclosure form information provided by the authors is available with the full text of this article at Neurology.org/cp.