Magnetic resonance imaging (MRI) is now widely accepted as an important diagnostic technology. Use of MRI to image blood vessels (magnetic resonance angiography) is increasing rapidly. From a rigorous review of the extensive published literature on MRI for neuroimaging, the American College of Physicians has updated its 1987 policy statement about MRI of the brain and spine.

Technical Advantages and Disadvantages of Magnetic Resonance Imaging

Three Advantages

Soft tissue contrast is better with MRI than other imaging techniques. Magnetic resonance images can be acquired with equal clarity in any orientation: axial, sagittal, coronal, or oblique. Image artifacts from bone are absent with MRI.

Magnetic resonance imaging does not involve exposure to ionizing radiation and has no known significant clinical side effects [3.1]. Magnetic resonance imaging without contrast can be done during pregnancy if necessary [3.1].

Magnetic resonance imaging does not require contrast in most situations, so it can replace computed tomography (CT) when the risk for an adverse reaction to iodinated contrast agents is high [3.1]. Paramagnetic contrast agents also are safer than iodinated contrast agents, although both usually are contraindicated in pregnancy [3.1].

Four Disadvantages

Magnetic resonance scanners are costly to install and operate [3.4].

Magnetic resonance imaging shows incidental anatomic abnormalities or anomalies that can be misinterpreted as causing a patient's symptoms [3.3]. Prominent examples are nonspecific white matter abnormalities [7.1, 12.2] or incidental spinal column abnormalities [13.6] whose prevalence increases with age.

Magnetic resonance imaging requires more patient cooperation than does CT [3.1]. Life support for medically unstable patients must be compatible with strong magnetic fields.

Metallic objects in the body may be contraindications for MRI. Most intracerebral aneurysm clips, intracranial or intraocular metal, and cochlear implants are absolute contraindications [3.1]. Pacemakers and other metallic implants usually are contraindications.

Principles of Assessment

This review assesses the diagnostic accuracy of MRI, its impact on therapeutic choices and on health outcomes of patients [2.2]. We reviewed the published literature [2.1], using previously defined methods for classification of the effect on patients and using ratings of methodologic quality [2.3, 2.4, 2.5]. The College recognizes that this approach generates conservative conclusions, especially when the technology has developed and spread rapidly into clinical practice.

The Quality of the Evidence

Since the previous review from the American College of Physicians, more information has become available about the sensitivity of MRI for a wider range of neurologic diseases and more is known about the occurrence of clinically silent findings on MRI. However, many studies continue to contain methodologic flaws that bias estimates of accuracy [16.1]. Problems with false-positive images or misinterpretations of inciden tal findings have not been studied [3.3, 16.5]. Only one study on the therapeutic impact of MRI used a randomized controlled trial design [2.3, 16.1, 16.3]. No studies of changes in patient outcomes have been published [16.1]. Studies of quality control or diagnostic accuracy for MRI in usual practice are not available. Overall, suggestions on when and how to use MRI in clinical practice must be based on weak evidence (Table 1).

Conclusions

General

Magnetic resonance imaging produces images of superb technical quality [4.1, 5.1, 6.1, 8.1, 9.1, 10.1, 12.1, 13.1 to 13.4, 15.1]. Magnetic resonance angiography and MRI have nearly equaled or exceeded the sensitiv ity of all competing technologies [4.1, 5.2, 9.2, 9.3, 12.2, 13.5, 14.1, 14.2]. Details of bony anatomy still require CT, and precise spatial resolution of blood vessels still requires conventional angiography [5.3]. Magnetic resonance imaging also shows more inciden tal anatomic abnormalities or anomalies of little clini cal significance that can be misinterpreted [3.3, 13.6, 16.5].

Changes in patient outcomes of mortality, disability, or pain relief have not been reported [16.1]. Like CT, MRI often documents abnormalities for which effective therapy is lacking. Magnetic resonance imaging can replace invasive diagnostic tests, such as myelography [13.7] and possibly carotid angiography [5.3].

Specific

Sudden Loss of Brain Function with or without Coma (Stroke)

Acute neurologic deficits, such as stroke, may require definitive investigation to determine the presence or absence of an intracranial hemorrhage or mass. Noncontrast CT scanning is sufficient for detection of acute intracranial bleeding and detection of most lesions large enough to cause increased intracranial pressure or mass effects [4.1, 6.1]. For small infarctions and most temporal lobe lesions, MRI gives the definitive diagnosis more often than does CT [4.1, 10.1]. Treatment planning for radiation or surgery is done with MRI if available. Magnetic resonance contrast agents may provide useful additional information in selected patients [3.2].

Suspected Carotid Artery Disease

Magnetic resonance angiography shows promise for noninvasive imaging of the cerebrovascular circulation [5.1]. Some centers are experienced, but published reports suggest wide variation in claims of accuracy [5.2]. Consensus on the appropriate use of magnetic resonance angiography with duplex ultrasound and conventional angiography must await completion of further studies [5.3, 5.4].

Dementia

Indications for MRI or CT in the assessment of dementia depend on the probability of detecting a reversible cause. Patients with rapid onset or progression of dementia, or with focal symptoms and signs, are candidates for imaging for possible reversible lesions. Because treatable symptoms derive from lesions that cause mass effects or hydrocephalus, CT with contrast or standard MRI is sufficient for most diagnosis [7.1]. For patients with stable dementia, imaging is not necessary [7.2].

Nonspecific white matter abnormalities are more common with advancing age and with increased risk factors for atherosclerotic vascular disease [7.1]. Alzheimer disease is associated with atrophy of the hippocampus and the temporal horns [7.1] without an increase in white matter abnormalities when compared with findings in age-matched control persons. Patients with vascular or mixed dementia have an increase in the number of such abnormalities [7.1]. Imaging adds little to the accuracy of clinical diagnoses, and prospective follow-up studies indicate that imaging has had no impact on treatment outcomes [7.2].

Cranial or Spinal Trauma

Magnetic resonance imaging shows injury to brain and spinal cord tissue better than any other technology [8.1, 13.2]. However, CT remains the study of choice for early, rapid screening for bony injury or intracranial bleeding. Magnetic resonance imaging findings are more highly correlated with successful functional recovery than is CT [8.1, 13.2], but studies showing improved selection of patients for treatment or rehabilitation or for improved patient outcomes are needed.

Suspected Brain Tumors or Cerebral Metastases

For imaging of suspected primary brain tumors or cerebral metastases, or for investigation of the brain while staging a known primary malignancy, MRI is better than CT. Early studies showed equivalent sensitivities, and further studies have shown that magnetic resonance contrast agents reliably separate metastases from incidental, benign, nonenhancing lesions of the white matter [9.2]. Poorly controlled studies suggest that MRI can influence treatment choices in about 20% of patients, but no formal patient outcome studies are available [9.4].

Suspected Pituitary and Juxtasellar Lesions

Based on early studies, MRI is the preferred imaging test for pituitary or juxtasellar lesions [9.3]. High resolution, thin-slice CT with contrast remains an acceptable imaging alternative if MRI is not available.

Epilepsy

Computed tomography with contrast is sufficient as an initial investigation in adults with new onset of seizures. Diagnostic yield is low unless seizures have a focal component. For complex partial seizures or seizures refractory to therapy, MRI shows more lesions than CT and may show a treatable underlying condition [10.1].

Suspected Cerebral Infection

Although studies of MRI for infections have been small case series that only showed image quality, MRI is now the preferred test for brain infections [11.1]. Computed tomography is sufficient for detection of mass lesions before a lumbar puncture. For patients with advanced human immunodeficiency virus (HIV) infection or the acquired immunodeficiency syndrome (AIDS), MRI gives more complete and definitive diagnoses of infections or tumor deposits than does CT [11.2]. Paramagnetic contrast agents can improve diagnosis in selected patients [11.3]. In any seriously ill patient, the preferred test is one that can be obtained expeditiously: either CT or MRI. No formal studies of therapeutic impact or patient outcomes have been published.

Suspected Multiple Sclerosis

Magnetic resonance imaging is the most sensitive imaging technique for detection of intracranial lesions associated with multiple sclerosis, but it also has the lowest specificity of all tests [12.2]. White matter abnormalities on cranial MRI indicate some increased risk for developing multiple sclerosis in patients with optic neuritis or idiopathic myelopathy [12.1]. Magnetic resonance imaging of the brainstem or spinal cord is preferred as the initial test when symptoms of multiple sclerosis mimic those of other brainstem or spinal cord lesions. No studies have defined therapeutic or patient outcome impact attributable to MRI among patients with either definite or suspected multiple sclerosis [12.3]. Serial imaging with MRI to monitor response or course of disease offers little or no benefit in clinical practice and should be restricted to formal clinical trials of new therapies [12.3].

Spinal Diseases

Magnetic resonance imaging shows excellent image quality for spinal column and cord diseases [13.1 to 13.4]. Patients with a history of spinal cord compression or of systemic malignancy with spine pain can have rapid, noninvasive, and accurate evaluations for spinal column metastasis with MRI [13.2]. Myelographic CT, plain CT, and MRI have equivalent diagnostic accuracy for herniated disk or spinal stenosis, but MRI does not require an intrathecal injection. These tests all are more accurate than plain myelography [13.5]. All imaging tests show asymptomatic herniated disks and stenoses that can complicate interpretation of positive imaging results [13.6]. Paramagnetic contrast agents are helpful when imaging the postoperative lumbar spine [13.4] or when defining spinal column tumors or infections.

Suspected Disease in the Posterior Fossa

Magnetic resonance imaging displays lesion boundaries and neighboring structures better with less image artifact than does CT [14.2]. For lesions such as hemorrhage or neoplasm, large enough to show mass effects or edema, noncontrast CT is as good as MRI for detection of the lesion. Magnetic resonance contrast agents improve detection of small tumors that have not distorted surrounding soft tissues. For tumors of the eighth cranial nerve, MRI is much more sensitive and noninvasive than is CT [9.2]. In a controlled trial of a randomized choice between CT or MRI as the first test in patients with symptoms of disease in the posterior fossa, fewer second tests were ordered if MRI was the first test, but treatment plans seldom changed after either second imaging test [14.3].

Other Possible Uses

In the evaluation of headache or dizziness, imaging by MRI or CT has not been useful [15.2] and should be reserved for the patient whose presentation indicates a focal problem. Applications of MRI to Parkinson disease, to hemifacial spasm, or to neuropsychiatric disorders may advance neuroanatomic understanding [15.1, 15.3], but no clinical management role for MRI in these conditions has been shown [15.3].

Research Methods

Since the 1986 survey, some good-quality clinical efficacy studies on MRI have been completed, but many more contain methodologic flaws that limit their clinical utility [16.1]. Rigorous definition of the reference standard by independent investigation of the anatomy and by careful clinical follow-up of all patients remains the cornerstone of a high-quality study [16.3]. Study designs must include assessments of the potential for misleading interpretations caused by artifacts or by imaging of incidental anatomic findings that are not the cause of clinical disease [16.5]. Studies also must define the ways that patients are selected so that readers can evaluate the applicability of the study results to their own practice setting.

Research Directions

Research for the next 5 years should focus on two questions [18.3]. When used in routine clinical practice, is the performance of MRI as good as is claimed by major development centers? When does MRI make a difference for appropriate treatment choices or for major patient outcomes? Although these studies are difficult and require collaboration with methodologically trained researchers from health services, they are essential to prove the value of MRI in clinical practice.

Health Policy

When costs and charges related to MRI are high, they constitute major barriers to more widespread acceptance of the technology [16.6]. Magnetic resonance imaging is most useful when it replaces other imaging technologies rather than when it supplements an already crowded arsenal of diagnostic weaponry. Efficient diagnostic strategies that follow selective rather than all-inclusive choices about imaging must be developed and their relative cost-effectiveness evaluated [16.6].

Based on the literature reviewed through March 1993, little evidence is available to define any impact of MRI on overall patient outcomes [16.1]. Improved outcomes depend on having effective treatments, and MRI is purely a diagnostic advance. Indirectly, through new ideas arising from new imaging insights into disease mechanisms, MRI may contribute to the development of improved therapies that do have a large impact on overall patient outcomes [12.3]. Therefore, health policies about funding of MRI should support its use in controlled clinical research settings.