EVIDENCE-BASED ANSWER
Patients exhibiting stroke symptoms should have brain imaging immediately within 3 hours of symptom onset (strength of recommendation [SOR]: A, based on systematic review). In the first 3 hours after a suspected cerebrovascular accident (CVA), noncontrast head computerized tomography (CT) is the gold standard for diagnosis of acute hemorrhagic stroke (SOR: C, based on expert panel consensus). However, the sensitivity for hemorrhage declines steeply 8 to 10 days after the event. Eligibility guidelines for acute thrombolytic therapy are currently based on use of CT to rule out acute hemorrhagic stroke.
Magnetic resonance imaging (MRI) may be equally accurate in diagnosing an acute hemorrhagic stroke if completed within 90 minutes of presentation for patients whose symptoms began fewer than 6 hours earlier (SOR: B, based on a single high-quality cohort study). MRI is more sensitive than CT for ischemic stroke in the first 24 hours of symptoms (SOR: B, based on systematic review of low-quality studies with consistent findings) and is more sensitive than CT in the diagnosis of hemorrhagic or ischemic stroke greater than 1 week after symptom onset (SOR: B, based on 1 high-quality prospective cohort study).
Evidence summary
The British National Health Service Health Technology Assessment (HTA) Programme published a systematic review on optimal brain imaging strategies for the diagnosis of stroke in July 2004.2 The HTA searched Medline and EMBASE from 1980 to 1999 and found 1903 studies relevant to diagnostic imaging for stroke. Only 25 studies reported the type of stroke diagnosed and the imaging reference standard. Thirteen of these 25 studies describe the time interval from symptom onset to imaging.
The HTA found a wide range of sensitivities for CT and MRI for both hemorrhagic and ischemic strokes at different time periods (TABLE) and noted that the quality of most of these studies was poor. Most of the studies identified were performed in academic stroke centers and had small sample sizes. Interpretation was masked in only 58% of studies. Few data were available on interobserver reliability, and neuroradiologists usually interpreted images. Two studies, totaling 165 patients, compared CT and MRI scans performed on the same day. However, the most “acute” time period reported was within 48 hours from symptom onset; neither study reported the order in which scans were performed, and only 1 study masked the neuroradiologist to the interpretation of the other modality.
After this systematic review, the HTA performed a prospective cohort study comparing CT with MRI obtained in random order on the day of presentation. They enrolled 228 patients presenting to a general hospital with stroke symptoms: (1) lasting longer than 1 day, but causing little or no decrease in function, or (2) lasting longer than 5 days. The mean time from onset of symptoms to scanning was 21.5 days. CT detected hemorrhagic stroke in 50%, and late hemorrhagic transformation in 20%, of those patients found to have hemorrhagic stroke on MRI, which was considered the criterion standard of chronic stroke diagnosis. The earliest hemorrhagic stroke missed by CT was 11 days old, and the latest hemorrhage correctly identified by CT was 14 days old.
An additional prospective cohort study comparing imaging modalities in the acute time frame has been published since the HTA review.2 The study enrolled 129 patients with stroke symptoms of less than 3 hours, as well as 71 patients with between 3 and 6 hours of symptoms. Patients underwent multimodal MRI (including gradient recalled echo and diffusion-weighted imaging) and noncontrast CT within 90 minutes of presentation. Two stroke specialists and 2 neuroradiologists, masked to clinical information, read the scans independently at a later date. Interrater reliability was good (= 0.75–0.94) for identifying acute hemorrhage. There was 96% concordance between the MRI and CT interpretations. The 4 hemorrhages “missed” by CT were hemorrhagic transformations of acute infarct, and the 4 hemorrhages “missed” by MRI were misclassified as chronic when they were acute.
The clinical implications of this study are uncertain. Without recognition of 1 imaging modality as the reference standard, it is difficult to say whether a hemorrhage “missed” on CT was a false negative on CT or a false positive on MRI.
TABLE
Range of sensitivities of CT and MRI for hemorrhagic and ischemic stroke
| NEUROIMAGING METHOD | TYPE OF STROKE | TIME SINCE SYMPTOM ONSET | |||
|---|---|---|---|---|---|
| >3 HOURS | >6 HOURS | <48 HOURS | >48 HOURS | ||
| Head CT without contrast | Sensitivity for hemorrhagic stroke | Evidence limited, assumed 100% in many studies | 86%–90%*3 | 93% | 17%–58% |
| Sensitivity for ischemic stroke | 64%–85% | 47%–80% | 23%–81% | 53%–74% | |
| Brain MRI | Sensitivity for hemorrhagic stroke | No MRI studies identified for this time frame | 86%–90%*3 | 46% | 38%–97% |
| Sensitivity for ischemic stroke | 65% | 84%–88% | 94%–98% | ||
| *Sensitivity=86% when other imaging modality used as gold standard, 90% when discharge diagnosis of acute hemorrhagic stroke used as gold standard. | |||||
| Source: Wardlaw et al, Health Technol Assess 20042; Kidwell et al, JAMA 2004.3 | |||||