|Home | About | Journals | Submit | Contact Us | Français|
Aneurysmal subarachnoid hemorrhage (SAH) is associated with high morbidity and mortality, and ischemic lesion burden is a known predictor of worse outcome. Currently, there is no single clinical method or ancillary test that can reliably predict which subset of patients will develop delayed cerebral ischemia (DCI). Previous animal and human studies suggest that SAH leads to a state of systemic inflammation, with DCI as the most striking manifestation. The aim of this study was to find hematological derangements and clinical factors present during the first seven days after bleeding that could help identify patients at risk for development of DCI.
Databank analysis of patients with SAH admitted between 2010–2012 in a large quaternary academic center. Data from demographics, imaging, laboratory and clinical factors were collected. Statistical testing was conducted to test for association to the outcome (DCI) and multivariate logistic regression was used to design a predictive model.
Of 55 patients, 14 developed DCI (25%). Anemia and leukocytosis on the third day after bleeding were significantly correlated with the outcome (p< 0.032, CI 1.12–15.16, OR 4.12 for anemia and p< 0.046, CI 1.03–26.13, OR 5.18 for leukocytosis). Anemia and leukocytosis were still statistically significant after adjustment for age, sex, modified Fisher scale and Hunt-Hess scale.
The presence of leukocytosis and anemia during the third day after SAH was statistically correlated with the occurrence of DCI. Further investigation is needed to assess the broader applicability of these findings.
Aneurysmal subarachnoid hemorrhage (SAH) is associated with high morbidity and mortality, with death rates ranging from 25% in contemporary observational series1 to close to 10% in randomized trials2,3. Roughly 30% of survivors suffer from severe permanent disability4, and significant cognitive and psychosocial impairment is present in a large amount of patients5–7. In patients with aneurysmal SAH, ischemic lesion burden is a known predictor of worse outcome8,9. Currently, there is no single clinical method or ancillary test that can reliably predict which subset of patients will develop delayed cerebral ischemia (DCI). The ability to stratify patients into high- and low-risk categories at the onset of hemorrhage would be useful in guiding early aggressive care. Transcranial doppler ultrasonography (TCD) has been used for decades as a tool for screening patients at risk of developing cerebral vasospasm, but the method lacks sensitivity and suffers from high interobserver variability10,11. Fisher and modified Fisher grades are based on the amount of subarachnoid blood on CT brain scan on admission and can fairly predict development of DCI, but with lower positive predictive values in patients with small amounts of blood12–15. Previous animal and human studies suggest that SAH leads to a state of systemic inflammation, with DCI as the most striking manifestation16–19. At the same time, anemia has been associated to worse outcomes and the occurrence of DCI20–23. Leukocytosis and anemia could represent early markers of bone marrow changes due to systemic inflammation, and serve as early surrogate markers for such a state. The aim of this study was to find hematological derangements and clinical factors present during the first seven days after bleeding that could help identify patients at risk for development of DCI.
A databank of aneurysmal subarachnoid hemorrhage was created using patients admitted from 2010 to 2012, in a large quaternary academic center in Cleveland, Ohio, USA. Fifty-five patients with aneurysmal SAH admitted through this period were included in our study. The entry criteria consisted of any patients with confirmed aneurysmal subarachnoid hemorrhage on cerebral angiogram, who stayed at least seven days in the hospital (counted from the day of bleeding) and with information available for all the study variables. We chose to analyze the first seven days after bleeding, as this is the peak period for occurrence of vasospasm in most patients with SAH. Characteristics of participant patients are presented in table 1.
Data from demographics (age, gender), imaging (transcranial doppler ultrasound, computed tomography and cerebral angiogram, including size and location of the aneurysm), laboratory (white blood cell count, hemoglobin concentration and serum sodium daily for the first seven days, cardiac dysfunction on echocardiography) and clinical factors (fever, occurrence of DCI and radiographic vasospasm, Hunt-Hess and modified Fisher scales on admission, surgical correction of the aneurysm) were collected.
We chose to analyze white blood cell count (WBC) and hemoglobin concentration on day three after bleeding, as this is usually when vasospasm ensues in most patients24,25. We defined anemia as a hemoglobin concentration of <10mg/dL, as this was found to be probably correlated with infarcts secondary to vasospasm and mortality in previous studies 21,22. Moreover, positron emission tomography studies in SAH patients indicate that raising hemoglobin concentration with red blood cell transfusion from 8 to 10 g/dL improves cerebral oxygen delivery26, and expert opinion per formal guidelines suggest that levels 8–10mg/dL should be maintained in patients with SAH, depending on the risk of delayed cerebral ischemia24. Leukocytosis is harder to define, as in theory it should be the total WBC count more than two standard deviations above the mean in a normally distributed population; most laboratories have values ranging from 10.000 to 11.000 leukocytes per mm3 as the upper limit for a WBC count27. We defined our cut-off as 10.000 leukocytes per mm3 after a receiver operating curve was done using the data from our study, correlating WBC with DCI, and it was found an area under the curve of 0.68, with values ranging from 10.01 to 10.46 x1000 leukocytes per mm3 to be suggested as with higher sensitivity/specificity. Youden index calculation of the best cutoff using the ROC curve values identified 10.03 as the most accurate value.
For statistical purposes, Hunt-Hess scale was dichotomized in 1–2 and >2 (good neurological grade versus disfavorable grade) and modified Fisher scale was dichotomized in 1–2 and 3–4 (lower risk of vasospasm versus higher risk).
We defined delayed cerebral ischemia as clinical deterioration deemed secondary to vasospasm (confirmed with digital subtraction angiography) after other causes were eliminated8 (such as fever, infection, hyponatremia, seizures, hydrocephalus) and radiographic vasospasm as arterial narrowing diagnosed on digital subtraction angiography24. Moderate narrowing was defined as at least 50% decrease in vessel lumen. In comatose patients, DCI was defined as new infarcts on brain computed tomography or magnetic resonance scans with radiographic vasospasm on digital subtraction angiography, and not deemed to be secondary to endovascular procedures24.
First, the chosen metabolic/clinical/radiologic variables were independently tested to check association to the outcome variable (DCI) using Pearson’s Chi-Square and Fisher’s exact test for categorical variables, t test for normally distributed continuous variables and Mann-Whitney U test for non-parametric continuous variables, as well as univariate logistic regression analysis. Afterwards, the variables with statistical significance (cutoff of p<0.05) were added to a multivariate logistic regression model to assess whether the statistical significance persisted after correction for pre-defined important demographic and clinical variables (Hunt-Hess, modified Fisher Scale, age and sex). Finally, a step-forward logistic regression modeling was performed to assess the best predictive model. All variables were independent, and no collinearity was detected. In the end, Hosmer-Lemeshow test was performed to assess goodness-of-fit of the final model. Statistical software BioStat® (version 5.4, Brazil) was used for all analyses.
Of 55 patients, 14 developed DCI (25%). Anemia and leukocytosis on the third day after bleeding were significantly correlated with the outcome (p< 0.032, CI 1.12–15.16, OR 4.12 for anemia and p< 0.046, CI 1.03–26.13, OR 5.18 for leukocytosis). The occurrence of fever or hyponatremia, age, sex, size and location of the aneurysm, surgical correction of the aneurysm, Hunt-Hess scale or modified Fisher scale on arrival, presence of cardiac dysfunction, as well middle cerebral artery mean flow velocities>120cm/s on transcranial doppler sonography were not correlated with the outcome. Table 2 shows the results for all tested variables. A multiple logistic regression was performed to assess the influence of the other non-significant variables when analyzed concurrently. Anemia and leukocytosis were still statistically significant after adjustment for age, sex, modified Fisher scale and Hunt-Hess scale in a logistic regression model using both variables (adjusted OR 6.3, p<0.019, CI 1.36–30.17 for anemia and adjusted OR 5.68, p<0.0367, CI 1.11–29.01 for leukocytosis). The median hemoglobin concentrations were 10.9 g/dL (SD 1.10) for the patients without DCI and 9.9 g/dL (SD 1.42) for the patients with DCI. Of note, no patients had a hemoglobin concentration of <10mg/dL on arrival. For the WBC count, a median value of 10.65 x1000/mm3 (SD 2.99) was found for the patients without DCI and 12.29 x1000/mm3 (SD 4.06) for the patients with DCI. Figures 1 and and22 show the hemoglobin and WBC counts graphed per population.
The findings remained the case even when retested including only patients with mean flow velocities of the middle cerebral arteries lower than 120cm/s on TCD sonography performed on the third day after bleeding. Even though our main goal was not to directly compare the diagnostic accuracy of TCD versus inflammatory markers, our data serves as a hypothesis generator that the presence of anemia and leukocytosis was predictive of DCI even among patients deemed of low risk according to TCD criteria.
In our study, the presence of hematological derangements on the third day after SAH was statistically correlated with the occurrence of DCI, even after correction for concurrent risk factors.
Some previous studies have noticed a trend for increased mortality and poor prognosis associated to leukocytosis during hospital stay in patients with aneurysmal SAH, but most analyzed the occurrence of blood count changes in any timeframe or even after DCI was established28–34. Also, the presence of leukocytosis was not corrected against other variables in regression models. One study conducted in the United States showed a relationship between peak leukocyte counts (above 15x 1000/mm3) and development of DCI, but it analyzed the presence of peaks above this cutoff occurring in any days between 0–5 post bleeding35. Moreover, it used a very high cutoff in a series were DCI developed in 45% of the patients, higher than our cohort and most previous studies24,36–38.
A growing number of animal and human studies suggest that SAH leads to a state of systemic inflammation, with vasospasm as the most striking manifestation16–19. Leukocytosis and anemia could be early markers of bone marrow changes due to systemic inflammation, and serve as early surrogate markers for this state. A cohort study in Sweden has shown that high leukocyte count at baseline was associated with increased incidence of SAH, mainly in smokers18. The third day after the initial hemorrhage is suggested to host the initial peak of inflammation in human and animal studies. Cerebral spinal fluid neutrophilia on the third day in patients with SAH was an independent predictor of the later development of DCI in a previous study39, and depletion of neutrophils three days after SAH in a murine model mitigated tissue inflammation, reversed cerebral vasoconstriction and restored memory mechanisms at day 640. Finally, none of the patients from our cohort developed infections during the seven-day period of observation, minimizing the possibility of observing leukocytosis secondary to bacterial infections.
Anemia has been associated to the occurrence of DCI and worse prognosis in previous studies21,41,42, as well as leading to metabolic distress detected by cerebral microdialysis20,23,43. Conversely, studies assessing the impact of blood transfusions in this population have led to conflicting results, not necessarily improving outcomes26,44–46. In critically ill patients, anemia is frequently observed and is hypothesized to occur as a consequence of systemic inflammation leading to myelosuppression, blunted erythropoietin production and erythropoietin response and abnormalities in iron metabolism47–49. It is quite common to detect anemia during the acute phase of SAH21,50,51, even as early as the 3–5th day after bleeding, before the effect of blood loss secondary to multiple blood draws is noticed, suggesting a possible involvement of systemic inflammation.
A point of discussion is that our study showed a statistical correlation, but causal inferences are harder to assert. For example, patients with infections or more frequent blood draws are likely sicker and might be in higher risk of developing DCI. However, in the present study blood samples were collected daily from all patients. Additionally, none of the patients received blood transfusions before day 3, and none were anemic on arrival. SAH is a condition where early diagnosis and aggressive therapy are some of the pillars of therapy, but until today very few data are available on predicting which subset of patients are at higher risk for complications. The modified Fisher and the Hunt-Hess scales are traditionally considered predictors of DCI, but were surprisingly not associated to the outcome in the current study. It might be related to its small sample size, but also to the fact that high scores in both scales could be closely related to the occurrence of early brain injury soon after aneurysm rupture52,53, impacting long-term functional outcomes even before DCI ensues.
Although the current study was retrospective and the cohort relatively small in number, it did disclose a trend for prediction of DCI. To our knowledge, it is the first in the literature to analyze in a very comprehensive way using multiple clinical, radiological and laboratorial variables the early phase of SAH, in regards of prediction of DCI occurrence. Studies on SAH are not easily performed, as the disease is not as common as other forms of acute brain disease; also studies analyzing the early phase soon after arrival of patients with SAH are scarce. We hope that our study could become a hypotheses generator for future investigation.
The presence of leukocytosis and anemia during the third day after subarachnoid hemorrhage was statistically correlated with the occurrence of DCI. Further investigation is needed to assess the broader applicability of these findings.