Magnetic Resonance Imaging and Computed Tomography in Emergency Assessment of Patients with Suspected Acute Stroke: a Prospective Comparison

Chalela JA, Kidwell C, Nentwich L, et al.
Lancet 2007;369:293–8

Background: The accuracy of multimodal MRI in assessing ischemic stroke and its impact on treatment and patient outcome is not clear. The authors compared the usefulness of CT and MRI for the assessment of acute stroke in the emergency setting.

Patients and Methods: Investigators at a US community hospital conducted this prospective study within 18 months. They included 356 consecutive patients with suspected acute stroke (median age 76 years, range 21–100 years), irrespective of symptom severity. 450 patients were screened and 94 were excluded – 49 because of MRI contraindications, and 45 because CT was not obtained or not interpretable. All patients underwent multimodal MRI (gradient echo imaging and diffusion-weighted imaging, performed after a median of 367 min [range 36 min to 8 days] after symptom onset) and noncontrast CT (performed after a median of 390 min [range 36 min to 8 days] after symptom onset) in a nonrandomized order. Four readers, who had no other patient information, interpreted the scans independently of each other. The diagnosis of ischemic stroke was given to all patients with imaging evidence of brain infarction, even if their deficits were transient. The main outcome variables were the accuracies of MRI and CT at diagnosing ischemic or hemorrhagic stroke.

Results: A total of 217 patients (61%) had a final diagnosis of acute stroke on the basis of clinical and imaging information; 190 strokes were ischemic (53% of the study population), with a median NIH Stroke Scale score of 3, and 27 strokes were hemorrhagic (8% of the study population). MRI was more sensitive than, and as specific as, CT at detecting any acute stroke (detection frequency 52 vs. 17%) and acute ischemic stroke (46 vs. 10%; p < 0.0001 for both comparisons). Among the 90 patients who were scanned within 3 h of symptom onset, MRI identified almost half of all cases with acute ischemic stroke, whereas CT identified less than 10% of those cases. MRI was also more sensitive (p < 0.0001) at detecting chronic intracranial hemorrhage, but the two imaging techniques did not differ significantly in terms of detecting acute intracranial hemorrhage. MRI had a sensitivity of 83% (95% CI 78–88%) and a specificity of 97% (95% CI 92–99%) for detecting any acute stroke, compared with a sensitivity of 26% (95% CI 20–32%) and a specificity of 98% (95% CI 93–99%) with CT. In the subset of patients who were scanned within 3 h of symptom onset, the respective sensitivity and specificity values were 76% (95% CI 64–86%) and 96% (95% CI 79–100%) with MRI, and 27% (95% CI 17–40%) and 100% (95% CI 85–100%) with CT.

Conclusion: MRI was more sensitive than, and as specific as, CT at detecting ischemic and hemorrhagic strokes within 3 h of symptom onset and beyond and should be the preferred test for guiding stroke treatment.

Comment

Diagnostic imaging in acute stroke can have an impact on different levels [1]. First of all, brain imaging will reduce health-care costs if it prevents stroke-related disability and death. Brain imaging can also improve the functional outcome after a stroke if it identifies patients who will benefit from an available treatment strategy, e.g., thrombolysis. To identify such patients and to enable the correct choice of treatment (therapeutic impact), brain imaging must provide relevant information. Relevant information is most likely the exclusion of stroke mimics and the demonstration of specific stroke pathology, e.g., for thrombolysis the exclusion of acute brain hemorrhage, the site and type of arterial obstruction, the extent of perfusion deficit, and the volume of ischemic damage (diagnostic impact). Moreover, the imaging modality should be accurate in detecting stroke pathology while treatment is still effective, e.g., within 3 h of stroke onset (diagnostic accuracy). To stay within this time limit, the imaging modality needs to be technically feasible in patients with acute stroke.

Chalela et al. examined a patient population with very mild strokes [2]. Half of the 190 patients with a final diagnosis of ischemic stroke presented with no or minor neurologic deficits. Patients with transient deficits but with imaging evidence of cerebral infarction were diagnosed as “ischemic strokes”. This means that the diagnosis of ischemic stroke in an unreported number of patients was based on positive MRI findings only, and these numbers were included in the sensitivity analysis. By no surprise, MRI has the best accuracy for strokes defined by MRI (circular reasoning). Moreover, 75% of the patients with presumed strokes were imaged beyond the accepted time window for thrombolytic therapy, some up to 8 days after stroke onset. The impact of imaging findings on treatment, clinical outcome and health-care costs was not assessed. Nevertheless, the authors suggest that MRI “might increase the cost-effectiveness of stroke care” and conclude that MRI “should be the preferred test for accurate diagnosis of patients with suspected acute stroke”. This is wishful thinking. The feasibility of MRI in patients with severe acute stroke is problematic [3], and the impact of MRI on treatment and clinical outcome of patients with ischemic stroke is not clear [4]. Chalela et al. forgot to mention that ischemic stroke can be effectively treated within the first 3 h of stroke onset on the basis of minimal imaging information, i.e., the exclusion of brain hemorrhage with CT. When comparing MRI with CT, it is important to bear in mind that the main requirements of stroke imaging tools are a high specificity for stroke pathology and a concept of how the imaging information can be translated into improved treatment and benefit for the patient. The high specificity of CT for ischemic damage [5], as confirmed in this study, makes the normal CT an important technique for the assessment of patients with acute stroke. Moreover, CT can easily be combined with CT angiography and CT perfusion imaging to show the territory of hypoperfusion.

The current article does not provide any evidence that MRI should be the preferred test in patients with acute stroke.

There is concern that unjustified conclusions from this study might cause stroke physicians to let their patients wait for MRI, thereby delaying subsequent treatment, while CT is readily available.

References

1. Fryback D, Thornbury J. The efficacy of diagnostic imaging. Med Decis Making 1991:88–94.
2. Chalela J, Kidwell C, Nentwich L, Luby M, Butmann J, Demchuk A, Hill M, et al. Magnetic resonance imaging and computed tomography in emergency assessment of patients with suspected acute stroke: a prospective comparison. Lancet 2007;369:293–8.
3. Hand P, Wardlaw J, Rowat A, Haisma J, Lindley R, Dennis M. MR brain imaging in patients with acute stroke – feasibility and patient-related difficulties. J Neurol Neurosurg Psychiatry 2005;76:1525–7.
4. Muir K, Buchan A, von Kummer R, Röther J, Baron JC. Imaging of acute stroke. Lancet Neurology 2006;5:755–68.
5. von Kummer R, Bourquain H, Bastianello S, Bozzao L, Manelfe C, Meier D, Hacke W. Early prediction of irreversible brain damage after ischemic stroke by computed tomography. Radiology 2001;219:95–100.

(submitted May 14, 2007)

Rüdiger von Kummer, Dresden