Magnetic resonance imaging in acute stroke

In this issue of Neurology are two articles describing further experience with magnetic resonance diffusion weighted imaging (DWI) and perfusion weighted imaging (PWI) in acute ischemic stroke.^1,2 Both add to the growing body of evidence from clinical observation and animal experiments documenting that these new MRI techniques show abnormalities in regions of acute cerebral ischemia earlier than either conventional MRI or CT. We would like to discuss the use of MRI as a diagnostic test for acute ischemic stroke in the context of the practice of clinical neurology. We will not discuss the use of MRI for clinical research or treatment trials because the requirements for a diagnostic test under these circumstances are quite different.

In the practice of medicine, the usual purpose of a diagnostic test is to improve the accuracy of the initial clinical diagnosis. The goal of improved diagnostic accuracy is not an end in itself, but a means to improve patient outcome by allowing selection of the most appropriate therapy. Thus, the value of a diagnostic test is determined not only by its accuracy but by demonstrating that the information it provides can be used for selecting the most effective patient management. Even in circumstances where no proven treatment exists, more accurate diagnosis can prevent the use of inappropriate treatment and provide prognostic information which is of value to patients and their families. A new diagnostic test is useful if it can replace a more invasive or expensive test or if it provides information that can be shown to improve outcome or reduce cost. Cost-benefit ratios calculated in terms of cost per quality-adjusted life years saved allow comparison of different diagnostic and treatment protocols.^3,4

Given these considerations, what is the role of MRI in the clinical evaluation of patients with suspected acute ischemic stroke? The current practice of clinical neurologic evaluation and noncontrast CT scan is 95% accurate in establishing the diagnosis.⁵ False-positive diagnoses primarily occur with seizures, toxic-metabolic encephalopathies, systemic infections, and peripheral neuropathies.^5,6 The diagnostic sensitivity and specificity of DWI and PWI in this setting are unknown. Experimental evidence of DWI changes with seizures and hypoglycemia suggest that it may be less than perfect.^7,8 False-negative diagnoses may occur, as documented by Tong et la.¹ A large, well-studied group of patients with suspected acute ischemic stroke is needed to establish the diagnostic accuracy of DWI. Such a study should be free of the defects in experimental design which have characterized much of diagnostic test research, such as failure to study patients with a wide spectrum of clinical characteristics and failure to mask interpreters of the test to patient information.^9,10 Even if the accuracy of DWI can be rigorously established within these guidelines, it is unlikely that its routine clinical use in patients with suspected acute ischemic stroke will have much impact because current diagnostic practices are already so accurate.

Tong et la.¹ provide data from patients with acute ischemic stroke showing a high correlation between the volume of both early DWI and PWI abnormalities to neurologic outcome, as measured by the 24-hour NIH stroke scale (NIHSS). These interesting data suggest a useful prognostic role for early MRI. However, there was substantial variation from patient to patient, including one patient with a major neurologic deficit at 24 hours who did not have a DWI lesion acutely (although PWI did show a large abnormality). Thus, there is some question about the accuracy of prediction in any individual case. In addition, patients with brainstem strokes were excluded. Inclusion of such strokes is likely to reduce the strength of the correlation between the size of image abnormality and the neurologic deficit. Most important was the finding that the relation between initial NIHSS and 24-hour NIHSS was excellent. This indicates that a simple bedside evaluation could provide the same outcome data as the sophisticated imaging procedure.

What about the use of MR imaging techniques to help guide treatment in individual patients? Although the benefits of r-tPA given within the first 3 hours to patients who meet the criteria used in the NIH trial are well established, some physicians are reluctant to give this drug to all eligible patients.^11,12 Advocates of thrombolytic therapy agree that some method for selecting individual patients who will benefit most would be an improvement.¹³ Recent subgroup analysis of the NIH r-tPA trial failed to provide such information.¹⁴ Conventional MRI has not been tested for evaluation of acute stroke patients, but because it is insensitive for delineating early ischemic changes or hemorrhages, it is unlikely to be of much value. As Tong et al. point out in their discussion, the use of combined PWI-DWI may be useful in the future for such a purpose because of its ability to provide physiologic and anatomic information.¹ Takano et al.² used measurements of the apparent diffusion coefficient and the perfusion index to compare the efficacy of intravenous versus intra-arterial thrombolytic therapy in an animal model of embolic stroke. However, the use of their technique for patient care was not evaluated.

The concept of imaging cerebral perfusion in patients with acute ischemia is not new. Many imaging methods older than DWI and PWI also show early changes in response to decreases in cerebral blood flow. These include single photon emission computed tomography (SPECT), positron emission tomography, and xenon inhalation regional cerebral blood flow methods. Several of these older methods provide accurate quantitative regional cerebral blood flow measurements. However, none of them became a standard for patient management. The relation between cerebral blood flow and tissue viability during acute ischemia is complicated. The ability of neurons to survive a period of cerebral ischemia depends not only on severity and duration but on a host of other factors.¹⁵ Even accurate quantitative measurements of regional cerebral blood flow have not proven useful for predicting tissue viability in acute ischemic stroke.¹⁶ There are no empiric data to demonstrate that measurement of cerebral blood flow is of value for making patient management decisions. Current clinical MR methods provide, at best, a qualitative index of tissue perfusion with an uncertain relation to quantitative measurements of cerebral blood flow. The application of new MR technology to guide therapy in acute ischemic stroke is enticing, but will require rigorous proof. A randomized trial should demonstrate that a group of patients, in whom the choice of therapy is based on the results of PWI-DWI, has a better outcome or lower cost than a similar group managed without such input. Such outcome-based trials of diagnostic tests have become the standard in the field.¹⁷

The relation between abnormal images and clinical outcomes is complex. We currently lack a gold standard to identify the full extent of brain tissue that is no longer functional or viable. Until such detailed understanding of the relations between tissue function and anatomic, physiologic, or biochemical abnormalities becomes available, it will not be possible to define damaged tissue using imaging technology precisely. However, it should be feasible to correlate images with clinical outcomes. The value of such information in clinical practice must be stringently defined. Cost-benefit analysis is a major feature of this process. Such a standard was not required in the past, but it has become essential, and simple assertions of the value of a new technique will no longer suffice. The Tong et al. and Takano et al. articles are part of the process that will be required to evaluate such promising new methods, but it is not yet clear whether PWI-DWI scans will become a standard of care for acute stroke victims.