Premorbid warfarin use and lower D-dimer levels are associated with a spontaneous early improvement in an atrial fibrillation-related stroke


Gyeong-Moon Kim, Department of Neurology, Samsung Medical Center, Sungkyunkwan University, 50 Irwon-dong, Gangnam-gu, Seoul 135-710, Korea.
Tel.: +82 2 3410 3598; fax: +82 2 3410 0052.

Spontaneous recanalization and early improvement have both been found to be more frequent and complete in patients with an embolic stroke when compared with other stroke subtypes [1]. The factors determining early recanalization and clinical improvement are very important in an embolic stroke because there are significant differences in prognosis that depend upon improvement during the early phase. Many cardioembolic stroke patients that do not clinically improve early develop worsening disease owing to poor collateral circulation secondary to the suddenness of the event [2]. We investigated clinical and laboratory markers as tools to predict a spontaneous early improvement in patients with atrial fibrillation (AF)-related cardioembolic stroke.

We analyzed data collected from patients admitted to Samsung Medical Center with an acute ischemic stroke between January 2005 and August 2010. Inclusion criteria were (i) patients admitted within the first 6 h after onset of symptoms; (ii) AF previously diagnosed, or detected during the admission period; (iii) an evident cardioembolic stroke according to the criteria outlined by the Stop Stroke Study Trial of Org 10172 in Acute Stroke Treatment (SSS-TOAST); and (iv) acute ischemic lesions on diffusion-weighted magnetic resonance imaging (MRI). We defined early improvement as a ≥ 4-point National Institutes of Health Stroke Scale (NIHSS) score improvement from the time of the initial evaluation to the time of the decision to thrombolyse [3]. Prehospital improvement was defined using the following items, which could be reliably collected retrospectively: motor paresis of the (i) arm, (ii) leg or (iii) facial muscles; (iv) aphasia, (v) dysarthria and (vi) sensory deficit. Only patients with a documented improvement in at least four criteria were evaluated further in this study [4].

Variables that could potentially affect the early improvement were recorded for each patient. A brain computerized tomography (CT) and CT angiography were performed upon arrival at our emergency room. Signs of a hyperdense middle cerebral artery (MCA) identified dense abnormal MCA attenuation compared with normal appearing contralateral MCA on unenhanced CT. Major arterial occlusion indicated clinically relevant arterial occlusion of the internal carotid, middle cerebral (M1 portion) or vertebrobasilar arteries.

We analyzed the differences between the early improvement and no improvement groups using the chi-square test or the Fisher’s exact test for categorical variables, and Student’s t-test or the Mann–Whitney U-test for continuous variables. Multivariable Cox regression analysis was used to identify independent predictors of early improvement.

Among the 101 patients who met the inclusion criteria, 20 patients (19.8%) showed spontaneous early improvement: 8 of these 20 patients showed prehospital improvement. The median NIHSS score improvement was six (interquartile range [IQR] 4–8). Comparisons of data from patients showing early improvement and those displaying no early improvement are included in Table S1. A history of a previous stroke was more common in patients in the early improvement group (30.0% vs. 8.6%, = 0.020). The frequency of premorbid warfarin use was higher in patients with an early improvement (60.0% vs. 4.9%, < 0.001). The median D-dimer level was significantly lower in the early improvement group than in the no early improvement group (0.33 vs. 1.07, < 0.001). On the basis of previous studies, the cut-off D-dimer level for this study was set at 0.5 μg mL−1 [5]. In the early improvement group, only 25.0% of patients showed elevated D-dimer levels of ≥ 0.5 μg mL−1 compared with 86.3% of patients in the no early improvement group. The median International Normalized Ratio (INR) level was higher (1.22 vs. 1.04, = 0.002) and the median C-reactive protein (CRP) level was lower in the early improvement group than in the no early improvement group (0.07 vs. 0.20, = 0.012).

A subgroup analysis of patients with premorbid warfarin use revealed a significant difference in median D-dimer levels between the early improvement and the no early improvement group (0.30 [IQR 0.24–0.55] vs. 0.85 [IQR 0.61–2.26], = 0.040). In contrast, there were no significant differences in INR levels (1.23 [IQR 1.22–1.80] vs. 1.50 [IQR 1.30–1.85], = 0.262).

Multivariable Cox regression included variables with < 0.1 in univariate analysis; initial NIHSS, a history of a previous stroke, premorbid warfarin use, D-dimer, INR and CRP (adjusted for time from onset to blood sampling). Premorbid warfarin use (odds ratio [OR]: 3.49, 95% confidence interval [CI]: 1.11–10.99) and a low D-dimer level (< 0.5 μg mL−1) (OR: 4.66, 95% CI: 1.40–15.5) remained independent predictors of an early improvement in an AF-related stroke. We divided the patients into four groups according to prior warfarin use and D-dimer levels. Our results indicate significant differences in the rate of early improvement (P < 0.001 for trend, Fig. 1). The area under the receiver-operating characteristic curve of premorbid warfarin use (+)/low D-dimer was 0.924 (95% CI: 0.843–1.000, < 0.001) for an early spontaneous improvement (Fig. 1). The positive predictive value of premorbid warfarin use (+)/low D-dimer level (< 0.5μg mL−1) was 88.9% (95% CI: 51.8–99.7) and the negative predictive value was 90.0% (95% CI: 81.2–95.6).

Figure 1.

 (A) The rate of early improvement according to warfarin use and D-dimer levels. (B) The area under the receiver-operating characteristic curve of premorbid warfarin use (+)/low D-dimer was 0.924.

We evaluated the relationship between these independent predictors and findings from the imaging to examine potential mechanisms underlying an early improvement. The prevalence of signs of hyperdense MCA and major arterial occlusion were significantly lower in the low D-dimer group than in the high D-dimer group (76.2% vs. 36.4%, = 0.001 and 40.9% vs. 67.2%, = 0.043, respectively). The prevalence of hyperdense MCA signs were significantly lower in the premorbid warfarin users than in the non-user group (26.7% vs. 70.7%, = 0.002).

Our study indicates that low D-dimer levels are associated with an early improvement in AF-related stroke. Increased D-dimer levels suggest on-going intravascular turnover and higher thrombotic burden [6]. Elevated D-dimer levels have been shown in patients with AF, especially in patients with documented thrombus formation within the left atrial appendage [7]. In our study, high D-dimer levels were associated with the prevalence of hyperdense MCA signs and major arterial occlusion. Currently available data, together with our results, suggest that increased D-dimer levels are indicators of a thrombotic burden. However, it is possible that increased D-dimer levels may reflect not only a high thrombotic burden, but also a large stroke volume as a consequence of a high thrombotic burden.

Prior warfarin therapy was another determinant of an early improvement in a cardioembolic stroke irrespective of the INR in our study. This finding suggests the possibility that prothrombin time is not the only valid measure of the effect of warfarin. A previous study showed a significant reduction in D-dimer levels in patients with AF after the introduction of warfarin, without a correlation between D-dimer and the INR [8]. They suggested that the INR is, therefore, not directly related to the efficacy of warfarin in reducing thrombogenesis, but rather a measure of the effect of warfarin on clotting factors (II, VII, IX and X), whereas the reduction in D-dimer reflects fibrin turnover and the efficacy of reducing the production and formation of fibrin. A recent study indicated that D-dimer levels predict subsequent thromboembolic events in patients with AF on anticoagulant therapy [5]. In our study, subgroup analysis of patients on premorbid warfarin revealed a significant difference in D-dimer levels between the early improvement and no improvement groups. Our results also suggested D-dimer levels might be a useful supplement to INR estimation in patients on anticoagulant therapy.

Our results have several limitations. This was a single center study examining a small cohort of patients. D-dimer values were missing in 12 (11.9%) of the 101 patients. We excluded the missing data after performing a test of missing completely at random. Even although data appeared to be MCAR in Little’s test, we could not completely exclude the risk of bias because of missing data. As patients with mild symptoms (NIHSS score < 4) were excluded by the definition of spontaneous early improvement, our study sample was not necessarily representative of all patients with an AF-related stroke. Validation in a large population is now needed.

Disclosure of Conflict of Interests

The authors state that they have no conflict of interest.