Cerebral Oxygen Transport Failure?: Decreasing Hemoglobin and Hematocrit Levels After Ischemic Stroke Predict Poor Outcome and Mortality
STroke: RelevAnt Impact of hemoGlobin, Hematocrit and Transfusion (STRAIGHT)—an Observational Study
Background and Purpose—Although conceivably relevant for penumbra oxygenation, the optimal levels of hemoglobin (Hb) and hematocrit (Hct) in patients with acute ischemic stroke are unknown.
Methods—We identified patients from our prospective local stroke database who received intravenous thrombolysis based on multimodal magnet resonance imaging during the years 1998 to 2009. A favorable outcome at 3 months was defined as a modified Rankin Scale score ≤2 and a poor outcome as a modified Rankin Scale score ≥3. The dynamics of Hemoglobin (Hb), Hematocrit (Hct), and other relevant laboratory parameters as well as cardiovascular risk factors were retrospectively assessed and analyzed between these 2 groups.
Results—Of 217 patients, 114 had a favorable and 103 a poor outcome. In a multivariable regression model, anemia until day 5 after admission (odds ratio [OR]=2.61; 95% CI, 1.33 to 5.11; P=0.005), Hb nadir (OR=0.81; 95% CI, 0.67 to 0.99; P=0.038), and Hct nadir (OR=0.93; 95% CI, 0.87 to 0.99; P=0.038) remained independent predictors for poor outcome at 3 months. Mortality after 3 months was independently associated with Hb nadir (OR=0.80; 95% CI, 0.65 to 0.98; P=0.028) and Hb decrease (OR=1.34; 95% CI, 1.01 to 1.76; P=0.04) as well as Hct decrease (OR=1.12; 95% CI, 1.01 to 1.23; P=0.027).
Conclusions—Poor outcome and mortality after ischemic stroke are strongly associated with low and further decreasing Hb and Hct levels. This decrease of Hb and Hct levels after admission might be more relevant and accessible to treatment than are baseline levels.
Sudden loss of the oxygen supply to brain tissue is, besides that of glucose, the first and central step in the pathophysiology of ischemic stroke. After the infarct core is rapidly and irreversibly destroyed, the penumbra and surrounding oligemic regions bridge the time to potential salvage by increasing oxygen extraction. Tissue fate depends on timely reperfusion or collateral perfusion, that is, regaining the oxygen supply. It therefore seems logical that an optimal oxygen-carrying capacity in the blood, that is, the level of hemoglobin (Hb), might be decisive for penumbral destiny. So far, the optimum Hb level in acute ischemic stroke has not been established, and definite guideline recommendations are lacking.1
Anemia (ie, Hb <12 g/dL in women and <13 g/dL in men, according to the World Health Organization definition2) is common in elderly patients, with a prevalence of ≈8.5% for those 65 years and older and is associated with increased mortality and hospitalization.3,4 A recent study demonstrated that approximately every fifth patient with acute ischemic stroke presents with anemia on admission and reported a close (“U-shaped”) relation between anemia on admission and poor outcome.5 Some studies demonstrated low Hb levels as a predictor of poor outcome after ischemic stroke,6,7 but other studies did not.8,9 Of note, all of those studies addressed Hb on admission only. However, it might be more relevant as to how Hb levels develop during the hospital stay, once ischemic stroke has occurred, as a putative relevance to outcome might turn Hb into an accessible treatment target.
Hematocrit (Hct) levels have been more actively studied in stroke patients than Hb levels, mainly with regard to the question of viscosity and cerebral blood flow. One study found an inverse relation between Hct and cerebral blood flow.10 Another described a higher baseline Hct value to be independently associated with infarct growth and a lesser chance of successful reperfusion after acute ischemic stroke.11 Several studies demonstrated a U-shaped relation between outcome parameters and Hct levels,9,12–14 suggesting an optimal midrange Hct level between 42% and 45%. Hemodilution as a therapeutic strategy in acute ischemic stroke, however, has not resulted in any clear benefit.15,16 Particularly in the Multicenter Austrian Hemodilution Stroke Trial, hemodilution did not significantly influence death and outcome parameters.17 Thus, hemodilution is not recommended as a therapeutic strategy in acute stroke, with the possible exception of patients with severe polycythemia.1,18
We performed an analysis of prospectively collected data on stroke patients to demonstrate not only the incidence and relevance of anemia on admission but also the association of further augmentation of anemia after admission with outcome and mortality. Several relevant hematologic and laboratory parameters, as well as clinical parameters and scores, were incorporated into the analysis to correct for potential confounders. Finally, we attempted to specify the types of anemia to suggest an approach to treatment.
From our local prospective database, we retrospectively analyzed all patients with acute ischemic stroke who received intravenous thrombolysis based on magnetic resonance imaging findings on admission during the years 1998 to 2009 (N=236). The selection of patients by these 2 criteria was chosen to provide a more homogeneous study population, comparable reperfusion times and strategies, and doubtless proof of acute ischemic stroke. To avoid heterogeneity, we excluded 19 patients who required treatment in our intensive care unit because of the much more complex and stronger influences on Hb and Hct dynamics, the difference in transfusion need caused by neurosurgical operations, and a different and stricter in-house transfusion protocol applied on our intensive care unit. For the remaining 217 patients, baseline and demographic characteristics, cardiovascular risk factors, and treatment time intervals were collected. Stroke severity on admission, as defined by the National Institutes of Health Stroke Scale (NIHSS) score, was assessed by physicians trained in the use of this scale. All patients were reassessed 90 days after stroke onset by either an outpatient visit or a telephone interview with the modified Rankin Scale (mRS) with a range from 0 to 6 (0=no symptoms, 6=death) by an investigator blinded to anemia status and laboratory data. Because patient independence was considered by us to be 1 of the most desirable treatment aims, a favorable outcome after 90 days was defined as an mRS ≤2 and a poor outcome as an mRS ≥3. As part of an additional sensitivity analysis, however, the definition of poor outcome as an mRS ≥4 was tested as well. All Hb and Hct concentrations during the hospital stay were extracted from the laboratory registry of the hospital. All patients had their first blood sample, including a blood count, taken in our Emergency Room. Intravenous thrombolysis was started in our Emergency Room and continued at our certified stroke unit, where patients typically stay for at least 3 days. Red blood cell (RBC) transfusion on our stroke unit is considered for all patients with an Hb <7 g/dL and at higher levels in individual situations. However, none of the patients included in this study received a RBC transfusion during their hospital stay. We assessed the levels of Hb and Hct on admission (baseline); mean, nadir, and maximum levels; and the maximal degree of Hb and Hct decrease from baseline, in addition to various other relevant laboratory parameters such as erythrocyte morphological parameters, blood glucose, leukocytes, C-reactive protein (CRP), creatinine, blood urea nitrogen, and sodium. To allow for the assumption that Hb/Hct dynamics might indeed have an impact on the penumbra and infarct development and not just reflect the general severity of illness apparent at a delayed time point during the hospital stay, we analyzed Hb/Hct dynamics within a rather close temporal relation to the insult, that is, during the first 5 days after admission. Anemia was defined by World Health Organization criteria as an Hb <12 g/dL in women and <13 g/dL in men. Elevated Hb levels were defined according to our in-house laboratory standards as Hb >15 g/dL in women and >17 g/dL in men. The normal range for Hct (36% to 47% in women and 38% to 52% in men), mean corpuscular volume (83 to 97 fL), mean corpuscular hemoglobin (27 to 33 pg), mean corpuscular Hb concentration (30 to 36 g/dL), leukocytes (4 to 10 nL−1), CRP (<5 mg/dL), creatinine (0.1 to 1.3 mg/dL), blood urea nitrogen (<45 mg/dL), and sodium (135 to 145 mmol/L) were defined according to our in-house laboratory standards. The study was approved by our local ethics committee (application No. S324/2009).
Distribution of the data was tested by using histograms and the 1-sample Kolmogorov-Smirnov test. For normally distributed data, the results are presented as mean±SD; for nonnormally distributed data, as median and interquartile range, or counts and percentages. Univariable analysis for comparison between patients with favorable and poor outcomes was performed by 1-way ANOVA as indicated. Because of the possible colinearity of tested variables, we applied a stepwise logistic-regression model. This model applies the variables one by one by using F statistics for the variable to be added exceeding the level of 0.05. After a variable is added, the procedure analyzes all included variables and deletes any variable that fails to produce an F value <0.1. After the necessary deletions are accomplished, another variable can be added to the model. The procedure terminates when no variable outside the model exceeds the necessary threshold and every variable included is significant. Hb and Hct parameters were tested separately in a multivariable analysis adjusted for age, NIHSS score on admission, and relevant potential laboratory confounders because of the close relation between these variables. The relation between Hb and Hct nadir levels and risk of poor outcome was assessed by the Kaplan-Meier estimator. For all statistical testing, we used the Statistical Package for the Social Science (SPSS Inc, 16.0 for Windows).
A total of 217 patients who were treated with magnetic resonance imaging–based intravenous thrombolysis during the years 1998 to 2009 were included in this analysis. Table 1 shows baseline characteristics and the distribution of outcomes. As expected, patients with poor outcomes were significantly older (64  versus 72  years, P<0.001) and had had more severe strokes than did patients with favorable outcomes (NIHSS score 10  versus 15 , P<0.001). Time to treatment (175  versus 170  minutes, P=0.37) and median length of hospital stay (6.5 versus 6.0 days, P=0.45) did not differ significantly between patients with a favorable and those with a poor outcome.
Altogether, anemia as defined by the World Health Organization definition was present in 33 (15.2%) patients on admission, in 86 (40.1%) until day 5 after admission, and in 93 (42.9%) until discharge. Anemia was about twice as frequent in patients with a poor outcome compared with those with a favorable outcome on admission (20.4% versus 10.5%, P=0.04) and within the first 5 days of the hospital stay (54.4% versus 27.2%, P<0.001). Lower baseline (14.1 versus 13.6 g/dL, P=0.016), mean (14.0 versus 13.2 g/dL, P<0.001), and nadir (13.3 versus 12.5 g/dL, P<0.001) Hb levels were significantly more frequent in patients with a poor than in those with a favorable outcome. Hb decrease was, on average, 0.9 g/dL in patients with a good outcome and 1.1 g/dL in patients with a poor outcome (P=0.011). An Hb decrease of >2.5 g/dL during the first 5 days after admission was more frequently associated with a poor than with a favorable outcome at a high level of significance (16.5% versus 5.3%, P=0.007). Elevated Hb levels were rarely observed and not significantly different between the 2 groups (6.1% versus 10.7%, P=0.23) of this population.
Lower baseline Hct (41% versus 40%, P=0.015), mean Hct (40.5% versus 38.5%, P<0.0014), nadir Hct (38% versus 36%, P<0.001), and maximum Hct (43% versus 41%, P=0.012) levels were significantly associated with a poor outcome. Reduced Hct levels were found in 48.3% of patients with a favorable outcome and in 77.7% of patients with a poor outcome (P<0.001). Elevated Hct levels were not observed in any of our patients (Table 1). When dichotomizing good versus poor outcome as an mRS of 0 to 3 versus 4 to 6, respectively, no relevant changes in the results regarding Hb and Hct were obtained.
Anemia and Clinical Laboratory Associations
During their hospital stay, 93 patients (43%) had anemia. Table 2 shows the distribution of baseline characteristics and cardiovascular risk factors between anemic and nonanemic patients. Anemic patients presented with more severe strokes, reflected by a median NIHSS score of 14 compared with 12 in nonanemic patients (P<0.001). Hypercholesterolemia was more often observed in nonanemic patients (31.5% versus 17.2%, P=0.017). In patients with anemia as opposed to those without anemia, RBCs were more often microcytic and hypochromic (17% [18.3%] versus 8% [6.5%], P=0.007; and 8% [8.6%] versus 2% [1.6%], P=0.015, respectively). There were no significant differences with regard to macrocytic and hyperchromic RBCs. Markers of inflammation and infection, that is, leukocytosis (64 [68.8%] versus 69 [55.6%], P=0.049) and CRP elevation (77 [82.8%] versus 88 [70.9%], P=0.044) were more frequent in anemic patients. Also, creatinine (17 [18.3%] versus 11 [8.9%], P=0.04) was significantly more often found to be higher and blood urea nitrogen (46 [49.4%] versus 46 [37.1%], P=0.07) to be trendwise higher in patients with anemia. With regard to hyponatremia or hypernatremia, there was no significant difference between the groups (Table 2).
Prediction of Outcome
In a multivariable regression model to predict poor outcome at 3 months including correction for age (odds ratio [OR]=1.04; 95% CI, 1.01 to 1.7; P=0.002), NIHSS score on admission (OR=1.23; 95% CI, 1.14 to 1.32; P<0.001), increased CRP (OR=2.7; 95% CI, 1.22 to 5.99; P=0.014), blood glucose (P=0.07), microcytic (P=0.70) and hypochromic (P=0.71) RBCs, leukocytosis (P=0.22), and increased creatinine (P=0.13), we found the presence of anemia (OR=2.61; 95% CI, 1.33 to 5.11; P=0.005), Hb nadir (OR=0.81; 95% CI, 0.67 to 0.99; P=0.038), and Hct nadir (OR=0.93; 95% CI, 0.87 to 0.99; P=0.038) until day 5 to be independently associated with poor outcome. There was no significant relation between baseline Hb (P=0.25) or baseline Hct (P=0.13) and poor outcome in our patients (Table 3).
When the multivariable analysis was performed for the complete hospital stay, that is, beyond the first 5 days, not only the aforementioned parameters but also Hb and Hct mean and Hb and Hct decrease showed a significant correlation with poor outcome (data not shown). Logistic regression for the poor outcome definition of mRS=4 to 6 demonstrated anemia until day 5, Hb nadir, and Hct nadir as independent predictors thereof.
Prediction of Mortality
Overall mortality after 3 months was 13.8%. In a multivariable logistic-regression model to predict mortality at 3 months adjusted to age (OR=1.06; 95% CI, 1.02 to 1.11; P=0.005), NIHSS score (P=0.08), blood glucose (P=0.22), microcytic (P=0.69) and hypochromic (P=0.35) RBCs, leukocytosis (P=0.09), increased creatinine (OR=2.71; 95% CI, 1.06 to 6.90; P=0.04), and increased CRP (P=0.22), we found Hb nadir (OR=0.80; 95% CI, 0.65 to 0.98; P=0.028), Hb decrease (OR=1.34; 95% CI, 1.01 to 1.76; P=0.04), and Hct decrease (OR=1.12; 95% CI, 1.01 to 1.23, P=0.027) to be independent predictors. NIHSS score on admission, anemia during the hospital stay, or mean and baseline values for Hb or Hct were not significantly associated with mortality after 3 months in our population (Table 4).
Of all Hb and Hct values, the nadir levels were the strongest predictors of poor outcome. The Figure demonstrates the risk of poor outcome in relation to Hb and Hct nadirs during the first 5 days after admission.
This study shows, for the first time, that developing or worsening anemia after admission for ischemic stroke predicts poor outcome and mortality. Anemia within the first 5 days after admission and the lowest levels of Hb and Hct independently predicted poor outcome (mRS >2) at 3 months. Hb decrease and nadir as well as Hct decrease independently predicted mortality at 3 months. Independent prediction prevailed despite adjustment for other reported predictors such as NIHSS score, age, CRP, glucose, or creatinine. Anemia or its indicators, Hb and Hct, are often associated with a greater severity of illness or organ failure and might be seen as reflectors of these conditions, which themselves have the potential to compromise stroke outcome. It is therefore important that clinical characteristics and laboratory parameters representative of these conditions were either not different between anemic and nonanemic patients or were entered into the multivariable analysis without changing the independent predictive potential of Hb and Hct decreases and nadirs. Thus, worsening anemia after ischemic stroke seems not just to reflect the sicker patients but could play an independent role in the pathophysiology of the fate of penumbral tissue, especially because we limited our analysis to the dynamics during the first 5 days after stroke. However, this retrospective analysis of clinical and laboratory data does not allow for more detailed assumptions on the mechanisms behind a potential causality.
The incidence of anemia in our selected intravenous thrombolysis–treated stroke patients was 15.2% and is in good agreement with a recently published study on admission Hb in ischemic stroke, reporting a rate of 19%.5 In contrast to that study, anemia on admission was not independently associated with poor outcome and mortality in our patients, possibly owing to our smaller population size. The focus of our work, however, was the further development of anemia after admission: the prevalence of anemia increased within the first 5 days of hospital stay to a rate of 40% in all patients and up to 54% in those with a poor outcome. In contrast to other studies,5,9,12–14 we did not observe a U-shaped but an inverse S-shaped relation between Hb or Hct nadir levels and poor outcome. According to this observation, the risk of a poor outcome was increased in those with an Hb level <13.5 g/dL and an Hct level <39% in our selected study population. However, we might not have observed a U-shaped relation because we had very few patients with elevated Hb or Hct levels in our population.
We tried to differentiate between certain types of anemia by assessing laboratory parameters that help to classify anemia or are often related to causes of anemia. Some of them were significantly higher in anemic than in nonanemic patients, but none of these parameters influenced the independent prediction of Hb and Hct levels for poor outcome and mortality. A higher incidence of microcytic and hypochromic RBCs in anemic patients possibly reflects iron incorporation disturbances or infection-related anemia. Increased creatinine and blood urea nitrogen might indicate renal insufficiency with secondarily impaired, erythropoietin-related erythrocyte generation. Leukocytosis and a raised CRP, often observed in our stroke patients, might reflect infection as a leading cause of anemia. Finally, reduced Hct and hyponatremia in anemic patients could be an indicator of iatrogenic fluid overload with hypervolemia as a common cause of anemia in hospitalized patients. Clearly, this subgroup analysis can only be a cautious speculative approach to explanation but does not allow for certainty or even weighting of anemia types and causes in this stroke population.
The clinical implications of our findings are not straightforward. Although they might seem to support a more active transfusion practice in stroke patients, the benefits of such measures, which carry the risk of several side effects, are doubtful and have yet to be proven in stroke patients. At present, our results should certainly not be understood as a call for more RBC transfusions in stroke treatment. So far, on the basis of our findings, we consider it reasonable to actively counteract anemia in stroke patients by limiting blood sampling to the necessary minimum, avoiding fluid overload, preventing and treating infections early, and controlling kidney function. Above all, it seems warranted to be aware of and look for anemia from the start, monitor it, and differentiate its types and causes by the respective diagnostic tests. We believe that the impact of anemia in patients with ischemic stroke is unjustifiably underestimated. More work is necessary on the pathophysiology, causes, forms, and treatment (such as RBC transfusion) of stroke-associated anemia. The aim should be to establish the optimum Hb for the stroke patient, if it exists.
Our study has limitations. Besides those of every retrospective analysis prone to selection and documentation bias, we might well have not taken unassessed confounders into account. It is very difficult to adjust for systemic severity of illness in these patients. Because a validated score is lacking, we took great care to address the severity of illness by including many individual clinical risk factors and laboratory parameters in our analyses. Although we believe that we have considerably minimized the possibility of anemia being a mere reflection of illness severity, we still cannot rule this out completely, as a retrospective analysis like this does not allow for adjustment for all conceivable influences. Because laboratory values were collected during a period of ≈10 years, methods of assessment might have changed. Only patients with magnetic resonance imaging–based thrombolysis were selected to achieve fair homogeneity among groups with regard to stroke age and recanalization strategy. This magnetic resonance imaging–based selection might be a reason why the median NIHSS score in our study was notably higher than in other stroke studies, for example, the ECASS 3 trial. Our findings might therefore not be translated without caution to subgroups with a different severity of ischemic stroke. In addition, this selection might have led to a population too small to detect a significant predictive effect of baseline Hb/Hct and other parameters.
Low and decreasing Hb and Hct levels are strongly associated with poor outcome and mortality after acute ischemic stroke. Further decreases in these 2 parameters during the hospital stay might be more relevant to the outcome than are the baseline admission levels and might be more accessible to treatment.
We like to thank Dr Robin Barrows for language editing.
- Received October 23, 2010.
- Accepted May 9, 2011.
- © 2011 American Heart Association, Inc.
- Adams HP,
- Del Zoppo G,
- Alberts MJ,
- Bhatt DL,
- Brass L,
- Furlan A,
- et al.
World Health Organization. Iron deficiency anemia assessment, prevention, and control: a guide for programme managers. http://whqlibdocwhoint/hq/2001/WHO_NHD_013pdf. Accessed October 23, 2010.
- Penninx BW,
- Pahor M,
- Woodman RC,
- Guralnik JM
- Wade JP,
- Taylor DW,
- Barnett HJ,
- Hachinksi VC
- Grotta J,
- Ackerman R,
- Correia J,
- Fallick G,
- Chang J
- Allport LE,
- Parsons MW,
- Butcher KS,
- MacGregor L,
- Desmond PM,
- Tress BM,
- et al
- Asplund K
- Aichner FT,
- Fazekas F,
- Brainin M,
- Polz W,
- Mamoli B,
- Zeiler K