Identification of "Tissue at Risk"
The use of thrombolytic agents is currently limited to patients who meet rigid time criteria, but a more rational approach to select patients for treatment and to study pathophysiology in individual patients is to use physiologic measurements that reflect the vascular and cellular pathobiology (34). Because it can measure physiologic variables in a short time, MRI can be used to investigate the pathologic features that best identify "tissue at risk" and clinical outcomes in response to therapies; these findings can be used for the design of clinical trials and for routine clinical care.
The use of DWI and PWI to make prognoses in individual patients is an active area of research (11,16,35,36). Data from natural history studies and clinical trials show significant correlations between the volume of abnormality on acute DWI and clinical severity; they also show that a combination of clinical factors (NIHSS score and time in hours from stroke onset to DWI) and DWI volume measurements is a better predictor of stroke outcome than the use of a single factor at acute or chronic times (11,12,16,17,19,37). Several studies have analyzed early MR characteristics that, in untreated stroke patients, predict the final infarct volume. We previously used a logistic regression model to differentiate regions of ultimate infarction versus noninfarction, based on baseline perfusion measures, and to operationally define the ischemic penumbra (36). Other groups have employed a variety of statistical and empirical approaches, such as generalized linear model algorithms, multiparametric ISODATA techniques, and other automated strategies, to predict final tissue outcome. In addition, thresholds for infarct progression and risk of hemorrhagic transformation (HT) might be identified by quantitative diffusion or perfusion MRI (36,38-42). However, these thresholds are not absolute and depend on the technique of measurement and analysis, the time from onset, the therapeutic intervention, and interactions with other physiologic and clinical variables (43-45). These and other approaches are accurate but are based on small and heterogeneous samples, and the validation, accuracy, reliability, and feasibility of these models in the clinic have not been reported.
Recurrent stroke is a major cause of morbidity and mortality, and the identification of effective interventions to prevent it is a research priority. The cumulative rate of clinical stroke recurrence within a few weeks after the initial ischemic event is < 5%, but as it is not easy to clinically differentiate worsening from a new ischemic event, it is probably that the true incidence rate is higher (46,47). Several groups have used DWI to study the prevalence of multiple acute ischemic lesions in patients with stroke and found that the lesions are common; their prevalence varies from 17% in patients imaged within 24 hr to 29% in the patients imaged within 4 days, and up to 83% in patients with high-grade carotid stenosis imaged within 1 week of onset (48-50).
To test the hypothesis that risk for recurrent ischemic lesions continued in the weeks following the clinically symptomatic stroke, we analyzed 2 cohorts of acute stroke patients. The first cohort involved 99 consecutive patients who had an MRI within 6 hr of symptom onset; 1 week after stroke, 34% had a recurrent infarct (in 15%, the new lesion was outside; in 16%, the new lesion was within the original area of perfusion abnormality; in 3%, a baseline PWI was not available). Initial multiple DWI lesions were associated with lesion recurrence (hazard ratio, HR, 2.8; 95% confidence interval (CI), 1.7-10.3; p=0.002) and with distant lesion recurrence
(HR, 6.0; 95% CI, 4.1-64.1; p < 0.0001). Large-artery atherosclerosis was the most frequent stroke subtype associated with lesion recurrence ( p=0.026) (51).
In a second cohort of 80 patients, the initial MRI was performed within 48 hr of onset and follow-up MRI was done at 5, 30, and 90 days; late lesion recurrence occurred in 26% of patients and was more frequently observed on the 30-day MRI than it was on the 90-day MRI (p=0.016). Early-lesion recurrence (HR, 3.0; 95% CI, 1.4-8.6; p=0.0095), distant early-lesion recurrence (HR, 3.6; 95% CI, 1.6-26.5; p=0.0088), and initial multiple DWI lesions correlating with early-lesion recurrence (HR, 3.6; 95% CI, 1.6-9.4; p=0.003) were associated with late-lesion recurrence. These early MRI markers and a history of hypertension remained significant, independent predictors of late-lesion recurrence. The frequency of early-lesion recurrence and late-lesion recurrence on MRI was significantly higher than that of early (2%) and late (4%) clinical recurrences ( p < 0.001) (52). These data suggest that, in some stroke patients, risk of recurrent ischemic lesions continues into the weeks following the clinically symptomatic stroke. If this is the case, MRI-defined ischemic lesion recurrence is potentially a useful surrogate endpoint in clinical trials of stroke prevention therapies. If studies establish the validity of this biomarker, it can be used in future clinical trials, requiring substantially fewer patients and shorter follow-up periods than do studies that exclusively rely on clinical outcome measures, with the ultimate benefit of enormous cost and time savings in evaluating preventive therapies.
The most effective therapy for acute stroke in humans is the recanalization of occluded arteries enabling tissue reperfusion (28,53). The effect of reperfusion, however, can lead to injury due to activation of the endothelium, excess production of oxygen free radicals, leukocyte recruitment, increases in cytokine production, enhanced inflammatory response, and edema formation. These changes damage the microvascular structure of the BBB, a prerequisite for HT (54,55). Several animal models focus on thrombolytic-related reperfusion injury and microvascular damage, and studies of animal models of stroke report parenchymal enhancement on MRI with BBB disruption (56-58). Until recently, the relevance of these findings to the pathology and treatment of acute stroke in humans was unknown. The discovery of a new MRI marker—hyperintense acute reperfusion marker (HARM)—allows the study of BBB disruption in stroke pathophysiology (59). HARM refers to delayed gadolinium enhancement of the CSF space on FLAIR that is due to disruption of the BBB (60). The BBB opening indicated by HARM is distinct from enhancement of the leptomeninges (61-63) and from the parenchymal enhancement that is observed during the later stage (days to weeks) of cerebral ischemia (Fig. 3) (61,64,65). In a recent study on a prospective cohort of 144 patients, we found HARM in 47 (33%), with a mean time from symptom onset to the observation of BBB disruption being 12.9 hr (SD = 10.3). However, based on the pharmacokinetics of gadolinium (66), it is likely that the opening of the BBB occurred before or soon after the administration of the contrast agent, at a median of 3.8 hr from onset. The timing and features of HARM are suggestive of the early BBB opening described in animals (67). The marker was focal in the sulcal space in the vascular territory of the acute stroke in 21 patients, diffuse within the ventricles in 6, and local and diffuse in 20. HARM was associated with reperfusion, subsequent HT, and poor clinical outcome (59). HT and BBB disruption were more common in patients treated with rtPA (31% and 55%) than in those not treated (14% and 25%; p=0.057 and 0.001, respectively). The association of HARM with HT was even greater in the subgroup of patients receiving rtPA, supporting the hypothesis that the proteolytic action of rtPA might contribute to extracellular matrix degradation and HT (68).
In collaboration with the University of California Los Angeles Stroke Center, we observed HARM in 62% of intra-arterial thrombolysis patients but only in 33% of patients treated with mechanical clot removal (69), suggesting that rapid reperfusion plays a causal role in early BBB disruption and that rtPA could exacerbate this condition. We believe that the pathophysiology indicated by HARM might present an opportunity for intervention prior to a failure of the BBB. Recent studies have implicated matrix metalloproteinases (MMPs), which attack the basal lamina and ultimately result in damage to the ultrastructure of the microvasculature (70 ). High plasma concentrations of MMP-9 are independent predictors of HT in treated and untreated patients. In the treated patients, however, it is unclear if high MMP levels are the direct result of the exogenous rtPA or reperfusion (71,72) . Early results using the MMP inhibitor BB-94 have shown a reduction in the hemorrhage rates, providing evidence that pharmacologic intervention might limit damage to the microvasculature and suggesting that MMP inhibition as a
Figure 3 Magnetic resonance images from a patient undergoing hemorrhagic transformation. (A) Aligned and registered diffusion-weighted imaging (DWI), fluid-attenuated inversion recovery (FLAIR), and gradient recalled echo (GRE) images acquired at 3 different times after onset of acute symptoms. DWI of initial examination 1 hr after onset depicts ischemic lesion as hyperintense. FLAIR images acquired 4 hr after onset exhibit evidence of blood-brain barrier disruption. GRE images at 28 hr show evidence of hemorrhagic transformation as a region of hyperintensity. Source: From Ref. 59.
promising target for reducing the hemorrhagic complications of thrombolytic therapy. HARM might have the potential as an imaging marker to evaluate those therapies. In that, as discussed above, the BBB is permeable close to the time window of acute thrombolytic therapy, this event is relevant to the development of therapies to prevent HT and improve outcome after throm-bolysis and other reperfusion therapies.
Was this article helpful?