Impact of Both Ends of the Hemoglobin Range on Clinical Outcomes in Acute Ischemic Stroke
Background and Purpose—Although both ends of the hemoglobin range may negatively influence clinical outcomes in acute ischemic stroke, most studies have examined the linear relationship or focused on the lower end of the range. Furthermore, it is unclear whether hemoglobin concentrations at different time points during hospitalization correlate with clinical outcomes in the same manner.
Methods—We identified 2681 consecutive patients with acute ischemic stroke from a prospective stroke registry database and grouped them into hemoglobin concentration quintiles using the following 5 indices: initial, nadir, time-averaged, discharge hemoglobin, and hemoglobin drop. To examine the effect of both ends of hemoglobin range, the third quintile was selected as a reference category except for hemoglobin drop, for which the first quintile was used. As outcome variables, 3-month modified Rankin Scale as an ordinal scale and 3-month mortality were used.
Results—With respect to higher modified Rankin Scale scores, the adjusted odds ratios and 95% confidence intervals of the first quintiles of initial, nadir, time-averaged, and discharge hemoglobin were 1.74 (1.31−2.31), 2.64 (2.09−3.33), 1.81 (1.42−2.30), and 1.65 (1.29−2.13), respectively. The opposite ends of these hemoglobin indices were not significantly associated. The adjusted odds ratio of the fifth quintile of hemoglobin drop (greatest hemoglobin drop) was 2.09 (1.51−2.89). The mortality analysis showed similar results except for initial hemoglobin.
Conclusions—In acute ischemic stroke, poor outcome was related to the lower but not the higher end of the hemoglobin range, regardless of when and how hemoglobin concentrations were measured.
A sudden interruption of oxygen to the brain is a crucial step in acute ischemic stroke (AIS). As red blood cells (RBCs) transport oxygen to the tissues and influence blood flow,1 RBC indices such as hemoglobin concentration or hematocrit could affect outcome of AIS.
However, the results of previous studies have been inconsistent,2–7 which may be explained, at least partly, by the hasty assumption regarding the shape of the relationship between RBC indices and clinical outcomes in AIS. Although both ends of the range of the RBC indices could negatively influence clinical outcome in AIS, most studies have examined the linear relationship or focused on the lower end of the range.3,5,6 Furthermore, RBC indices measured at different time points of hospitalization may behave differently with respect to clinical outcomes,6 but most studies only dealt with the RBC indices obtained at the time of admission.2–5,7
Here we report the impact of both ends of the hemoglobin range on clinical outcomes in AIS using hemoglobin indices measured at different time points of hospitalization.
A consecutive series of patients who were admitted to Seoul National University Bundang Hospital for AIS within 7 days of onset between January 2004 and November 2009 were identified using a prospective stroke registry. Modified Rankin Scale (mRS) score and mortality at 3 months were used as outcome variables.
We used the following 5 indices of hemoglobin concentration: initial, nadir, time-averaged, discharge hemoglobin, and hemoglobin drop (see online-only Data Supplement for definitions). Patients were categorized into quintiles according to each hemoglobin concentration. The third quintile was selected as a reference category except for hemoglobin drop, for which the first quintile was used. To examine the association between each quintile group and clinical outcomes, ordinal logistic regression analysis for a mRS score or binary logistic regression analysis for mortality was performed. Covariates whose associations with the initial hemoglobin were clinically relevant and whose P values for those associations were <0.2 were selected for adjustment.
The hemoglobin indices other than initial hemoglobin required ≥2 measurements of hemoglobin concentration during hospitalization. We used the sequential regression multiple imputation method to impute missing values of hemoglobin indices if hemoglobin concentration was measured only once.8 A complete case analysis was also performed as sensitivity analysis, and its results were also presented. Further details on the study methods are provided in the online-only Data Supplement.
The study population consisted of 2681 patients whose hemoglobin concentration was measured once (n=891) or more (n=1790) during hospitalization (Figure I in the online-only Data Supplement). Baseline characteristics were summarized (Table I in the online-only Data Supplement) and compared according to the initial hemoglobin quintiles (Table II in the online-only Data Supplement).
With respect to the initial hemoglobin, the odds of higher mRS scores at 3 months significantly increased only in the first quintile (Table III in the online-only Data Supplement). The adjusted odds ratio (OR) of the first quintile was 1.74 with a 95% confidence interval of 1.31 to 2.31 (Figure 1). With respect to the nadir, time-averaged, and discharge hemoglobin values, the adjusted ORs for increments of mRS scores also significantly increased in the first quintiles but not in the remaining quintiles (Figure 2). For the hemoglobin drop, the adjusted OR for increments of mRS scores elevated only in the fifth quintile.
As for mortality, significant increases in the adjusted ORs were observed in the first quintiles of the initial (Figure 1), nadir, time-averaged, and discharge hemoglobin (Figure 3); no significant changes were observed in the remaining quintiles except for the initial hemoglobin (Table III in the online-only Data Supplement). Notably, the adjusted ORs of initial hemoglobin for mortality significantly increased in the fourth and fifth quintiles. With respect to the hemoglobin drop, the OR for mortality significantly increased in the fifth quintile only.
The results of complete case analysis using the proportional odds models (Table IV in the online-only Data Supplement) and partial proportional odds models (Table V in the online-only Data Supplement), and confining the analysis to subjects hospitalized within 48 hours of onset (Table VI in the online-only Data Supplement), showed significant associations for higher mRS scores (Figure II in the online-only Data Supplement) and mortality (Figure III in the online-only Data Supplement) with the first quintiles of the nadir, time-averaged, and discharge hemoglobin as well as with the fifth quintile of the hemoglobin drop, similar to those of sequential regression multiple imputation analysis. Inclusion of blood pressure drop (decrease of >30% from the baseline in mean arterial blood pressure) in the logistic regression did not change the results (data not shown).
This study demonstrates that worse functional outcome measured by mRS and mortality 3 months after stroke were related to the lower but not the higher end of the hemoglobin range. These nonlinear relationships were consistent regardless of when and how hemoglobin concentrations were measured, which may contradict a recent report that highlighted the importance of the nadir hemoglobin.6
The multiple imputation technique in this study included many replacements of missing values using models based on the data. Therefore, the uncertainty of such replacements remains. However, because excluding cases or variables with missing values from analysis may lead to bias, the multiple imputation analysis is considered a better alternative to simply abandoning observed data due to missing values.8
The putative association between hemoglobin concentration and clinical outcome might indicate that hemoglobin concentration is a useful treatment target. Indeed, transfusion of hemoglobin polymers actually reduced infarct volume in an animal model.9 Furthermore, blood transfusion to patients with acute myocardial infarction and anemia decreased mortality.10 However, the effect of transfusion in anemic patients with AIS has not been evaluated to date and should be examined in future studies.
Sources of Funding
This study was supported by grants from the Korea Healthcare Technology Research & Development Project, Ministry of Health and Welfare, Republic of Korea (A102065).
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.113.002672/-/DC1.
- Received June 30, 2013.
- Revision received July 26, 2013.
- Accepted July 30, 2013.
- © 2013 American Heart Association, Inc.
- Wood JH,
- Kee DB Jr.
- LaRue L,
- Alter M,
- Lai SM,
- Friday G,
- Sobel E,
- Levitt L,
- et al
- Kellert L,
- Martin E,
- Sykora M,
- Bauer H,
- Gussmann P,
- Diedler J,
- et al
- He Y
- Mito T,
- Nemoto M,
- Kwansa H,
- Sampei K,
- Habeeb M,
- Murphy SJ,
- et al