However, a second CT or MRI study is often obtained later, even if the first study suggests a diagnosis. Several observations are cited as justification for serial neuroimaging studies. For example, both CT and MRI are more sensitive several days after an acute ischemic stroke than in the initial 24 hours after symptoms begin. Therefore, an additional study, usually an MRI, is often obtained if the initial study is normal. Physicians may also seek serial neuroimaging studies to confirm a diagnosis, exclude additional lesions not seen on the first study, and better define a lesion seen initially [5]. Evidence suggests that MRI may be more accurate than CT scanning in detecting lacunae, multiple lesions, and lesions in the posterior fossa [4, 6, 7]. Finally, an additional study may be needed because the patient's condition has worsened or, if anticoagulation is desired later in a patient's hospital course, to determine whether a stroke has become hemorrhagic.

Despite the widespread practice of obtaining serial neuroimaging scans after acute ischemic stroke, no studies have examined the effect of these scans on patient outcome and only one study has examined their influence on patient management [7]. Shuaib and colleagues [8] postulated that obtaining an MRI scan after an initial CT scan for ischemic stroke affected both diagnostic classification and certainty as well as therapy. However, this study had several limitations, including the absence of outcome data, procurement of additional neuroimaging studies to guide therapy before ancillary studies were examined, and an ambiguous description of the manner in which MRI objectively resulted in better localization of infarction.

For a series of 206 patients with acute stroke, we compared differences in stroke classification, therapy, and outcome between patients who had one neuroimaging study and those who had more than one.

Methods

Top Methods Results Discussion Author & Article Info References

Patients

We retrospectively reviewed the charts of all patients admitted to Madigan Army Medical Center with the diagnosis of acute ischemic stroke (International Classification of Diseases, Ninth Revision diagnosis 436 [disease, cerebrovascular, acute]) from 1990 to 1993. We excluded patients if they were younger than age 18 years, had a neurologic deficit lasting less than 24 hours, or had an admitting diagnosis other than acute stroke. All patients had a new or worsened neurologic deficit on initial physical examination. It is standard practice at our institution to obtain a noncontrast CT or MRI study of the head before or at the time of admission. We also excluded patients if radiographic findings strongly suggested intracranial processes other than stroke. These conditions included subarachnoid hemorrhage, subdural hematoma, epidural hematoma, brain abscess, and hydrocephalus. If the radiologist believed acute stroke could not be differentiated from one of the exclusionary categories, we included the patient in our analysis.

Setting

Madigan Army Medical Center is a 415-bed tertiary care medical center that offers residencies in internal medicine, neurology, and family practice. All admitted patients were cared for on the internal medicine, neurology, or family practice wards.

Stroke Classification Protocol

We retrospectively classified strokes on the basis of the physical examination at presentation, initial neuroimaging study, and all ancillary studies before a second CT or MRI study, if any, was obtained. Ancillary studies included echocardiography, carotid duplex ultrasonography, transcranial duplex ultrasonography, chest roentgenography, electrocardiography, and cerebral angiography. Ancillary studies and serial neuroimaging studies were ordered at the discretion of the attending physician. The classification scheme was derived from the Trial of Org 10172 (a low-molecular-weight heparinoid) in Acute Stroke Therapy (TOAST) because of the trial's rigid inclusion criteria and published low incidence of intraobserver variability in classification [9]. Our TOAST-based classification scheme is shown in Table 1. We further classified the strokes as probable or possible according to the clinical certainty of the diagnosis. We reclassified the strokes if additional neuroimaging studies were obtained.

Documentation of Therapy and Outcome

We used the physician order sheet to determine the type and timing of therapy. We also documented the time of initiation of therapy in relation to neuroimaging studies and any changes in therapy after neuroimaging studies were obtained. Interventions included antiplatelet therapy with aspirin or ticlopidine, anticoagulation with heparin or warfarin, carotid endarterectomy, and medication for vasculitis. We recorded patient outcomes at discharge and divided them into three categories: death in the hospital; discharge to the patient's home without professional home health services; or transfer to a skilled nursing facility or professional home health care service with requirements for nursing, physical therapy, or aides because of inability to perform activities of daily living.

Neuroimaging Studies

The CT scanner used was a General Electric (Milwaukee, Wisconsin) 9800 Highlite Advantage. Slices were 5 mm thick in the posterior fossa and 10 mm thick at higher levels. All initial CT scans were obtained without contrast, but all CT scans obtained as second or third neuroimaging studies were done with and without iodinated contrast. The MRI studies were done using a 1.5-Tesla magnet (Signa, General Electric) with 4.8 software. Five sets of images were routinely obtained: sagittal T1-weighted (TR [relaxation time]/TE [echo time]/NEX [number of excitation] equals 416/11/1), axial T1-weighted, axial T2-weighted (2200/30-80/0.5), and axial and coronal T1-weighted scans done after the administration of gadolinium diethylenetriaminepentoacetic acid. Matrices of 256 x 256 steps, a 20- to 24-cm field view, and 1 to 2 averages were used for the acquisitions. Slices were 5 mm thick with 2.5-mm slice spacing.

Statistical Analysis

Data are presented as the mean ±SD or as a percentage of the group value from which they were derived. We examined overall differences between groups by the Student t-test for continuous variables and chi-square tests for categorical variables. Differences were considered statistically significant at P < 0.05.

Results

Top Methods Results Discussion Author & Article Info References

Patient Characteristics

Most patients (92%) were older than age 60 years (mean age, 66.0 ± 10.8 years). Male patients composed about two thirds of the study sample. Most patients (82.5%) were white, 9.2% were black, and 8.3% were Asian. Almost all patients (94%) had one or more risk factors for stroke, and some (15%) had previously had a stroke. The characteristics of patients who had one neuroimaging study and of those who had two or more were similar, with no significant differences in any category (P > 0.1) (Table 2).

Number and Timing of Neuroimaging Studies

All patients had at least one CT or MRI study. Most studies (95.6%) were noncontrast CT scans obtained a median of 24.2 hours after onset of symptoms. About two thirds of the patients (68%) had two or more neuroimaging studies. Most of these studies (79.5%) were MRI scans obtained a median of 4.8 days after onset of symptoms; the rest were contrast-enhanced CT scans. Only 7.3% of patients had three neuroimaging studies, and 2.4% had four. We obtained the reasons for any additional studies from the patient chart and radiology request form. More than half of the studies (51.4%) were obtained to search for the cause of a stroke when classification was difficult. Approximately 38% of the serial studies were ordered to better define the anatomy of a stroke or to increase the confidence in the diagnosis of a stroke for which an initial diagnosis had already been made. Only 8.6% of studies were obtained because of a decline in neurologic status. A few studies (2.1%) were obtained for other reasons, such as to ensure that a bland stroke had not become hemorrhagic before anticoagulant therapy was started.

Initial Stroke Classification

We initially classified strokes according to the TOAST criteria based on the results of the physical examination at presentation, the first neuroimaging study, and results of any ancillary tests done before a second neuroimaging study, if any, was obtained. The most common type of stroke overall was small-vessel (38.3%), followed by strokes of unknown cause (32.5%). Large-vessel atherosclerosis, cardioembolism, and other causes composed 18.5%, 10.2%, and 0.5% of the diagnoses, respectively. In the patients who had multiple neuroimaging studies, initial classification into the unknown cause category accounted for more than half of cases (53.6%) compared with only 35.4% of cases in patients having only one imaging study.

Effect of Serial Neuroimaging Studies on Stroke Classification

Among the 140 patients who had two or more neuroimaging studies, 20% of strokes were reclassified according to the TOAST criteria into a different stroke category as a result of the additional study (Table 3). All reclassified strokes were switched from the unknown cause category to a category with a known cause. We reclassified no strokes initially placed in a category with a known cause. However, the certainty factor increased from possible to probable in 32.3% of persons with strokes of known cause who had two or more neuroimaging studies. All such strokes were in the small-vessel category and were found in patients with classic lacunar syndromes in which a lacuna was not present on the initial CT or MRI study but was seen on a subsequent study. Of note, in 20 of these 21 cases (95%), MRI detected the initially unseen lacuna.

Nature and Timing of Therapy

Most patients (78.2%) received a change in medication or were referred for carotid endarterectomy during their hospital stay. Fewer patients (16.0%) continued to receive their outpatient preadmission doses of anticoagulant or antiplatelet therapy without modification, and fewer still (5.9%) received no therapy at all. Changes in therapy consisted primarily of starting or increasing doses of antiplatelet agents (66.5%), starting or adjusting doses of anticoagulant agents (9.2%), or recommending carotid endarterectomy (1.9%). One patient was prescribed steroids for vasculitis. Aspirin was much more widely used for antiplatelet therapy than ticlopidine, accounting for 97.1% of antiplatelet medications compared with 2.9% for ticlopidine. In 93.2% of the patients whose medications were changed or who were referred for carotid endarterectomy, therapy was begun after the results of the first neuroimaging study were known. In only 6.8% of patients receiving new treatment was therapy initiated or altered after a subsequent CT or MRI study. Of these patients, most (75%) had strokes of initially unknown cause.

Patient Outcome

We measured short-term patient outcome at the end of each patient's hospital stay. Of the patients who had one neuroimaging study, 70.1% went home, 24.0% went to a skilled nursing facility or required daily home nursing or home health aide care, and 5.9% died; the corresponding percentages for patients who had multiple studies were 73.3%, 24.4%, and 2.2% (P > 0.1).

Duration of Hospital Stay

Duration of hospital stay was 10.33 ± 10.38 days for patients who had one study and 12.27 ± 9.16 days for patients who had multiple studies; this difference was not statistically significant (P > 0.1)

Discussion

Top Methods Results Discussion Author & Article Info References

The results of many previous studies of diagnostic instruments and therapy for acute ischemic stroke are difficult to interpret because of ambiguous classification systems. Intraobserver variation has been high unless rigid stroke classification criteria were applied. The TOAST classification system is a novel method that uses easily applicable clinical criteria to classify a stroke with low intraobserver variability. However, the rigid classification criteria in some cases put the strokes of patients with suggestive neurologic or historical findings in the unknown cause category if the patients did not have confirmatory findings on diagnostic testing. We believe that although this made our unknown cause category larger than that in many stroke studies, the strokes we could categorize were less apt to be misclassified. This hypothesis is supported by the fact that although most patients had serial neuroimaging studies, we reclassified no stroke previously classified as having a known cause.

Our results suggest that serial neuroimaging studies have a role in diagnosing the mechanism of strokes of unknown cause. With a second CT or MRI study, we found a likely cause of stroke in nearly 40% of strokes initially classified as having an unknown cause. Although evidence suggests that MRI is more sensitive than CT scanning for detecting lesions (especially lacunae) several days after onset of stroke, our study was not designed to determine whether one type of imaging was superior to the other.

For the strokes of 12 patients in whom CT or MRI studies were used to evaluate a deterioration in the patient's condition, the diagnostic classification did not change. However, these studies affected management decisions in 58.3% of the patients and provided valuable prognostic data. Of interest, no patient with a small-vessel stroke who presented more than 24 hours after symptoms developed had clinical deterioration or a serial neuroimaging study to investigate that possibility.

One group of patients that had additional neuroimaging studies deserves special mention. Most patients for whom the cause of stroke was known and who had an additional CT or MRI study had a stroke in the small-vessel category. All of these patients with suspected small-vessel disease had a second CT or MRI study to better define or detect a lacuna. The initial study found a supportive culprit lacuna in only 44.7% of the patients. Of the 49 neuroimaging studies ordered to define a lacuna not visualized on an initial CT scan, 50.0% of the 40 MRI scans found the lacuna compared with 11.1% of the 9 CT scans. Although a suggestive lacuna was found with an additional study in 42.8% of cases, diagnosis or therapy was not changed for any patient.

Lacunae are often difficult to find at any time after a stroke, and MRI is most sensitive 2 to 3 weeks after onset of symptoms [4]. With the low likelihood of clinical deterioration and the greater sensitivity of a later MRI for detecting lesions, a single MRI done several days after presentation in a patient with a stable classic lacunar syndrome and no history of cardiac or carotid disease may be an acceptable alternative.

Of the 78.1% of patients whose stroke-related therapy was changed, most (64.1%) had antiplatelet therapy initiated or increased. In only two of these patients was therapy withheld, for unknown reasons, until after a second neuroimaging study had been obtained. Anticoagulation was the second most common type of therapy used, most often for cardioembolic strokes, large-vessel "strokes in evolution," or strokes for which antiplatelet therapy had previously failed. Of the 12 patients in whom therapy was withheld or changed pending the result of an additional neuroimaging study, 9 (75%) had strokes of unknown cause in which optimal treatment was uncertain; 3 (25%) had large-vessel or cardioembolic strokes for which anticoagulation or admission to the intensive care unit was contemplated on the basis of interval changes seen on imaging. The patients received anticoagulation or immunosuppressive vasculitis therapy or had carotid endarterectomy after the results of the second study clarified the situation. In no case did a third or fourth study affect therapy.

Surprisingly, some patients (5.8%) received no stroke-related therapy. Furthermore, in a small but significant proportion of patients (13.1%), outpatient preadmission antiplatelet or anticoagulant regimens were not increased or changed, despite a new neurologic event. Possible explanations include that fact that data may have been obtained from a period during which anticoagulation for cardioembolic stroke and the use of ticlopidine were more controversial than they are now, inclusion of chronically debilitated patients for whom medications such as warfarin or ticlopidine may have provided more risks than benefits, and a need for continuing medical education in the latest techniques of stroke management in cases in which initial antiplatelet therapy failed [10].

No significant differences were seen in patient disposition or duration of hospital stay between patients who had one neuroimaging study and those who had more than one. Although CT scanning and MRI are diagnostic and not therapeutic imaging techniques, we believed it was important to show that limiting a stable patient with a likely diagnosis to one neuroimaging study would not affect patient outcome.

Our study has several limitations that may impede generalization of the findings. First, the TOAST criteria are one of many stroke classification systems, none of which has gained widespread acceptance outside of research circles. The effect of neuroimaging studies may differ when a diagnosis of stroke is less tightly or consistently controlled. It also may be difficult for clinicians to obtain ancillary tests, such as carotid duplex ultrasonography or echocardiography, in a timely manner outside a tertiary care center to aid in precise diagnostic categorization. However, we believe that as progress is made in devising therapies for acute stroke that are based on cause, stroke classifications schemes such as the TOAST system will become more clinically useful.

Multiple neuroimaging studies may be useful in determining the cause of acute ischemic stroke when classification is not obvious after the first neuroimaging study and ancillary studies have been done. They have no diagnostic and little therapeutic effect in stable patients in whom an initial diagnosis can be made with firm inclusion criteria. An additional CT or MRI study is most useful for choosing a therapy when a stroke is initially classified as having an unknown cause, when the stroke is particularly severe, or when neurologic deterioration is suggested. Patient outcome does not seem to be adversely affected in patients who have one neuroimaging study after stroke.

Current Author Addressees: Dr. Stephen M. Salerno: Department of Medicine, Madigan Army Medical Center, ATTN: MCHJ-M, Tacoma, WA 98431-5000.

Dr. Landry: Department of Medicine (MED-EDP), Uniformed Services University of the Health Sciences, 4301 Jones Bridge Road, Bethesda, MD 20814.

Dr. Schick: Division of Neurology, Landstuhl Regional Army Medical Center, CMR 402 Box 1335, APO AE 09180.

Dr. Schoomaker: Graduate Medical Education, HQ MEDCOM, 2050 Worth Road, Ft. Sam Houston, TX 78234-6000.