1. Magnetic resonance imaging (MRI) has revolutionized neuroimaging. Many internists now are familiar with the technology. However, strong support for a greatly expanded role of MRI for diagnosis and management of neurologic diseases is lacking because the evidence for clinical efficacy in the published literature is incomplete [1-3]. Because many more studies have appeared in the past 6 years, the American College of Physicians decided to update its 1987 report on the clinical efficacy of MRI in neuroimaging [4]. This review provides the basis for the updated assessment and accompanying suggestions for use of MRI in various clinical situations.

2. Methods

2.1 Titles and abstracts of 3125 citations from the peer-reviewed medical literature from 1987 through November 1993 were found using MEDLINE and ancillary searches. After screening out reviews, technical reports, case reports, and other publications describing fewer than 30 original case-patients, we identified 285 papers for further review. Two investigators independently read and classified each article according to the disease studied, the applicable level of clinical efficacy, and the methodologic quality of the study design. This three-part classification is summarized in Table 1 and has been described in detail previously [3]. Discrepant ratings were resolved by consensus.

2.2 Because the imaging capabilities of MRI have been widely discussed in case series, we focused on the 156 articles that addressed diagnostic accuracy or diagnostic impact, therapeutic impact, or patient outcomes. For diagnostic accuracy and effect, methodologic quality was rated as A if the study had more than 35 patients with and more than 35 without the pathologic abnormality in question, drawn from a clinically relevant sample whose clinical symptoms were completely described, whose diagnoses were defined by an appropriate reference standard, and whose magnetic resonance images were technically of high quality and were evaluated independently of the reference diagnosis. Studies were rated D if they had no credible reference standard for diagnosis, or if the test result and determination of final diagnosis were not independent, or if the sample size was smaller than 35 for patients with and without the disease, or if the source of the patient cohort could not be determined or was obviously influenced by the test results (work-up bias).

2.3 Studies of therapeutic impact or changes in patient outcomes were evaluated for methodologic quality according to a similar A through D scale, adapted from Sackett [5]. For example, an A-rated study required a randomized controlled trial using a clinically relevant, well-described sample, with adequate power to detect differences in well-characterized outcomes. The rationale and details for these ratings have been described [3, 6].

2.4 Among 143 articles reporting on diagnostic accuracy or impact, 1 was rated A, 14 were rated B, 14 were rated C, and 113 were rated D for methodologic quality. Deficiencies included small numbers of patients, failure to describe the source of patients and their spectrum of disease, failure to maintain independence between interpretation of the MRI scan and an appropriate reference standard, or obsolescence of the magnetic resonance technology. Estimates of sensitivity and specificity were derived from the A, B, and C studies. Results from receiver-operating characteristic (ROC) curves are discussed in the text when available. Among 21 articles reporting on the therapeutic impact of MRI, none were rated A, 1 randomized trial was rated B, and 2 comparisons with computed tomography (CT) were rated C (Table 2). Five studies comparing treatment plans before and after MRI, but without comparison to an alternative test, were rated D for methodologic quality. Retrospective surveys of anecdotal experiences were not rated. No studies provided data on long-term patient outcomes.

2.5 Suggestions for practice are given in the companion report from the American College of Physicians [4]. A strong "strength-of-evidence" rating would have been given to a suggestion supported by randomized controlled studies of patient outcomes and by conclusive evidence about diagnostic accuracy. The one suggestion given a moderate rating was supported by studies that used comparison groups to show improved therapeutic choices and a definite impact on diagnoses. Most of the suggestions given weak ratings were supported by fair or good studies of diagnostic accuracy but had no supporting studies of therapeutic impact or patient outcomes. The suggestion given an equivocal rating was based only on prevalent expert opinion and several reports describing a predominance of clinical practice patterns.

3. Safety, Technical Issues, and Costs

3.1 Safety evaluations indicate that noncontrast MRI is safe for most patients, including pregnant women [7-9]. Pacemakers, many cerebral aneurysm clips, intraocular metal, and cochlear implants are definite contraindications to MRI. Other vascular clips, vena cava filters, and most metallic implants require individual patient consultation [10]. Incomplete or technically inadequate studies occur because of patient or equipment problems in 3% to 10% of attempted scans [11-13]. By comparison, safety concerns about CT are focused on iodinated contrast dye reactions. Some acute reaction occurs in 3% to 12% of patients, with severe reactions in 0.02% to 0.2% of patients [14, 15]. Newer nonionic contrast agents are associated with lower reaction risks.

3.2 Paramagnetic contrast agents improve confidence in diagnosis in 45% to 75% of scans and actually change the apparent diagnosis in up to 30% of patients [12, 16-22]. None of these studies could report diagnostic accuracy statistics because the comparisons were done without any independent reference standards. The larger multicenter evaluations of contrast agents probably contain case-selection biases that skew statistical analysis [21]. Paramagnetic contrast agents cross the placenta and typically are contraindicated during pregnancy [9], but all are otherwise safe [23, 24]. Minor adverse reactions to the agents occur in 2% to 3% of patients, and anaphylaxis occurs in about 1 in 100 000 patients [25].

3.3 Diagnostic specificity for MRI has been studied only infrequently [3]. Although "specificity" has been used to mean that an imaging test can determine "specific" tissue composition, specificity as used in this paper is one minus the false-positive rate [26, 27]. Classically, a false-positive result is a positive test result when no disease is present. However, MRI often shows incidental anatomic abnormalities that are real and technically are not false-positive images (examples in Table 3. If any incidental abnormality of the cerebral white matter is called a false-positive result, the apparent diagnostic specificity can be low (0.28 to 0.58 [28-31]). The specificity is much higher if a more stringent test interpretation, such as one requiring multiple abnormalities to indicate disease, is used (0.80 to 0.90 [28]). The reliability of readings about white matter abnormalities is low [32]. No studies have reported investigations into the incidence of therapeutic misadventures caused by overinterpretation of white matter findings or by other asymptomatic abnormalities.

3.4 Costs of MRI historically have been high compared with competing technologies. Facility and professional charges for MRI are 30% to 100% higher than for CT (Table 4). Some payers reimburse MRI only at a rate equivalent to CT (for example, Medi-Cal). If MRI obviates further tests but CT does not, using MRI as the initial test may be less expensive than using CT. Similarly, if MRI replaces an invasive test associated with inpatient facility charges (such as a myelogram, cisternogram, or cerebral arteriogram), then the MRI cost is less than the total cost of the alternative. New fast-imaging sequences, contrast agents, and studies focused only on high-yield views for particular diagnoses can shorten examination times and improve efficiency at imaging centers.

4. Stroke and Transient Ischemic Attacks

4.1 The clinical diagnosis of the cause of a stroke is erroneous in 10% to 20% of patients [33], so most patients have imaging. Because spatial resolution of MRI has improved, smaller infarctions have been seen, often on the first day after onset of illness [34-36]. Among several studies with small sample sizes [13, 34, 37-39], the sensitivity of MRI for the infarction ranged from 0.86 to 0.95, whereas the sensitivity of CT ranged from 0.50 to 0.68. One study of posterior circulation infarctions also reported a specificity of 0.88 [38].

4.2 No studies rated as using grade C or better methods have compared MRI with either noncontrast or contrast-enhanced CT. Further, the impact of MRI on final diagnosis or therapeutic choices for patients with stroke has been reported only once [40]. In that study, MRI reportedly led to a change in diagnosis of the cause of stroke for 16% of patients and to changes in anticoagulation management for 19%. Treatment changes were not tabulated or evaluated for appropriateness; biasing of case selection through the influence of referral or nonreferral for imaging by treating clinicians also was not considered. No studies have investigated changes in patient outcomes attributable to the use of MRI after stroke.

5. Carotid Artery Disease and Magnetic Resonance Angiography

5.1 New techniques for magnetic resonance angiography are developing rapidly [41]. Between 5% and 10% of procedures fail for technical reasons [42, 43]. Accuracy in magnetic resonance angiography varies widely because equipment, technical skills, and interpretation skills also vary across centers [44].

5.2 Diagnostic Accuracy

Diagnosing a vessel as occluded or mildly stenosed when a severe stenosis is present is a false-negative error because in either case surgery might not occur. Similarly, interpreting mild stenoses or occlusions as moderate or severe stenoses is a false-positive error leading to unnecessary surgery [42]. In recent studies with B or C ratings for methodologic quality, the sensitivity of magnetic resonance angiography for angiographically significant stenoses ranged from 0.86 to 1.0 Table 5 [42, 43, 45-51]. Stenosis by magnetic resonance angiography correlated exactly with conventional angiography in 50% to 90% of arteries reported. Specificity varied from 0.60 to 0.98 (Table 5).

5.3 Diagnostic Impact

Although a combination of duplex and magnetic resonance angiography has been suggested [43, 52, 53] as a replacement for conventional carotid angiography, recent reviews [42, 54] concluded that magnetic resonance angiography was not yet accurate enough to replace conventional carotid angiography. The latter carries a risk for stroke or death ranging from 0.1% to 3% [55]. In one major center where patients received both treatments, investigators thought that 69% of patients could have been managed without the conventional angiogram [56]. Further, clinical information and duplex ultrasound results alone may be sufficient for appropriate management decisions in 87% of patients [57]. Most centers continue to use conventional angiography before doing an operation [47, 58].

5.4 Therapeutic Impact

No methodologically satisfactory studies have yet reported data showing that magnetic resonance angiography changes therapeutic choices or improves patient outcomes.

6. Intracranial Hemorrhage and Aneurysms

6.1 Studies of MRI for intracranial hemorrhages (whether intracerebral, subarachnoid, subdural, or epidural) have been technical reports and case series with small sample sizes [59-62]. When acute subarachnoid hemorrhage is suspected, a noncontrast CT is recommended, followed by a lumbar puncture for spinal fluid analysis [52, 63].

6.2 Magnetic resonance angiography and MRI may improve detection of small intracranial aneurysms [64-66], but spatial resolution is not adequate for planning of aneurysm surgery [67]. Leading centers have reported initial success doing stereotactic radiosurgery guided by MRI for vascular malformations [68, 69], but the therapeutic impact or patient outcomes for hemorrhage or aneurysm have not been studied rigorously.

7. Dementia

7.1 Interpretation of incidental abnormalities of white matter and clinically silent infarctions seen on MRI in older people is problematic. The prevalence of abnormalities increases from 20% among persons younger than 50 years to 90% for those older than 70 years; the prevalence doubles or triples within each age group when hypertension, diabetes, or previous cardiovascular disease is present [70, 71]. White matter abnormalities generally correlate with cognitive decline on formal testing, but long-term follow-up of healthy elderly patients has shown that many people with white matter abnormalities have no loss of cognition [72, 73]. Correlations between abnormalities in the hippocampus and Alzheimer dementia recently were reported [74-77]. Although these studies used control groups and rigorous analytic procedures, other studies [78] suggest that hippocampal atrophy is common among healthy elderly persons (Table 3).

7.2 No studies of the accuracy of MRI alone compared with CT for detection of treatable lesions in demented patients have been published. Older studies of CT as a diagnostic adjunct for dementia work-ups have shown little impact [79-82]. Fewer than 5% of patients with dementia have reversible causes and more than half are diagnosed by serum chemistry results rather than by imaging [81]. In a prospective study of 210 demented patients, 12% had a potentially treatable lesion discovered by CT or MRI. Only 4 had treatment, and none improved as a result [83].

8. Brain Injury and Head Trauma

8.1 Magnetic resonance imaging shows numerous parenchymal abnormalities after head injury [84-87], including unsuspected abnormalities in 10% of patients with minor head injury [88]. In these studies, CT detected fewer than half the abnormalities shown on MRI. However, CT findings reliably predict patients who need intensive neurosurgical management [89, 90]. In a recent case series, surgery was indicated by MRI alone for only 1 of 155 patients [91]. Certain MRI findings are correlated with better or worse recovery [92-94], but no studies have analyzed the additional contribution of these findings to clinical predictions of recovery and rehabilitation needs. Research into these management and prognosis questions remains underdeveloped [95].

9. Brain Neoplasms

9.1 Patients with metastatic or primary brain tumors may show progressive neurologic symptoms, have seizures, or be asymptomatic. Once a tumor is diagnosed, most centers use MRI to define its margins and effects on adjacent structures. However, infiltrating primary brain tumors often extend beyond MRI and CT image abnormalities [96, 97]. Computed tomography and MRI give high-quality images for meningiomas [98, 99], whereas contrast-enhanced MRI shows more anatomic details [98, 100]. Magnetic resonance imaging shows cerebellopontine angle and eighth cranial nerve tumors better than CT [101-103], although false-positive images do occur [104]. Enhancement of meningeal metastases with paramagnetic contrast is not reliable enough for definite diagnosis [105-108].

9.2 Diagnostic Accuracy and Impact

Few studies have evaluated diagnostic accuracy in a case series of patients suspected of having brain tumors. In older studies, CT and MRI had equivalent sensitivities for large lesions [2, 109]. In a recent study [110], MRI detected more presumed metastatic lesions than did CT, but independent confirmation of the lesions was not obtained. One study [111] (rated as using grade C methods) indicated that lesions seen on MRI, which were not enhanced by paramagnetic contrast, were not metastases.

9.3 Magnetic resonance imaging is more sensitive than high-resolution contrast CT for detection of sellar or juxtasellar lesions [2, 112, 113]. Lesion margins are better defined by MRI [113, 114]. With either test, false-positive results may occur because of clinically silent pituitary abnormalities.

9.4 Therapeutic Impact and Patient Outcomes

Only a few small studies examined the impact of MRI on treatment planning for brain neoplasms [115-117]. In one follow-up study, only 1 of 45 patients had tumors reappear outside the original treatment volume defined by CT [118]. In an early comparison of treatment planning according to CT or MRI, neither test was superior [119]. In planning treatment volumes for radiation therapy, the measurement variation between CT- and MRI-derived volumes was no greater than the inter-observer variation in such measurements [120].

10. Epilepsy

10.1 In patients with epilepsy, abnormal results on MRI in retrospective case series have varied from 21% to 71% in patients [121, 122]. One grade C study [123] and several grade D studies [121, 124-126] indicate that MRI yields more anatomic information than does CT. Paramagnetic contrast agents may change 5% of diagnoses [35]. Magnetic resonance imaging improves diagnosis when the history, the physical examination, or the electroencephalogram suggests a focal lesion [127].

10.2 In the only study [128] reporting apparent false-positive findings on MRI, the investigators attributed the errors to incomplete tissue sampling at surgery rather than to misinterpretations of the MRI scans. Some asymmetry in temporal lobe structures is normal [129] and may be a source of misleading images. Recent studies rated as using grade B and C methods indicate that abnormalities or atrophy of the hippocampus seen on MRI correlates with abnormal pathologic findings [128] and predict the location of the seizure focus [130-133].

11. Infectious Diseases

11.1 Because central nervous system infections are uncommon, collection of enough patients for a valid prospective study of diagnostic accuracy is difficult. However, MRI is the study of choice for imaging infections and inflammations of the spine [134, 135]. Most recent studies investigated spinal or brain infections related to the acquired immunodeficiency syndrome (AIDS).

11.2 Human Immunodeficiency Virus and the Acquired Immunodeficiency Syndrome

Imaging of primary encephalopathy caused by human immunodeficiency virus (HIV) has been imprecise [136]. Abnormalities of the central nervous system are seen in 12% of asymptomatic patients with HIV, no higher than in healthy concurrent controls [137] or in controls from other studies [138]. Clinical studies indicate that cognitive deficits among asymptomatic patients with HIV are no more common than in noninfected control patients [139, 140].

11.3 Although MRI is considered the most sensitive imaging modality for detection of infections and neoplastic complications of advanced HIV and the acquired immunodeficiency syndrome, no studies rated as using grade C or better methods have been published. Studies claiming high sensitivity for MRI [141] or changes in treatment plans [142] did not describe case-recruitment protocols that would ensure an unbiased selection of study patients. Recent expert opinion indicates that paramagnetic contrast agents should be reserved for unusual lesions because few diagnoses and fewer treatment plans would be altered by the routine use of these agents [143, 144]. No study has reported improved patient outcomes.

12. Multiple Sclerosis and Related Syndromes

12.1 Studies [145, 146] report clinically definite multiple sclerosis developing after optic neuritis or idiopathic myelopathy in 13% to 85% of patients. Variations in follow-up, case-finding, and definitions of multiple sclerosis affect these estimates [147]. Reliable conclusions about MRI findings and progression to clinically definite multiple sclerosis require follow-up of more patients in methodologically sound studies [140, 146, 148-150].

12.2 Diagnostic Accuracy and Impact

Magnetic resonance imaging shows white matter findings typical of multiple sclerosis in 70% to 95% of clinically definite cases, and reported specificity ranges from 0.75 to 1.0 [2, 151-155]. When rigorously evaluated, interpretations of MRI that gave sensitivities of 0.60 or 0.80 for multiple sclerosis were associated with specificities of 0.75 and 0.91, respectively, depending on definitions of image abnormalities [156] (Table 6). The problem of false-positive findings increases with age, especially for persons older than 50 years [155]. Use of MRI may have little overall diagnostic effect: In a census area of 250 000 people, only 3 of 72 new diagnoses of multiple sclerosis would have been delayed without MRI [157].

12.3 Therapeutic Impact and Patient Outcomes

Magnetic resonance imaging has had little direct impact on therapeutics or patient outcomes for multiple sclerosis. Although an abnormal MRI might impair a person's ability to get insurance coverage, patients with unexplained symptoms often appreciate the clarification of their illness offered by MRI [158]. Recent reviews have focused on the information value of MRI for the clinician-scientist rather than for the clinical practitioner [159, 160]. Methods for using MRI in longitudinal clinical trials are under study [161-165]. If MRI allows more rapid evaluation of new therapies, its value to patient care through research may be considerable [166].

13. Spinal Diseases

13.1 Magnetic resonance imaging has become the preferred modality for imaging spinal diseases ranging from radiculopathies to malignancies to spinal cord injury [167-171]. For spinal column metastases and spinal cord compression, MRI requires no intrathecal injection, images the entire spine rather than just the spinal cord, and gives more precise imaging of metastases [172-176]. For detecting and managing a syrinx, MRI is the most comprehensive imaging modality [177-179]. Only six studies of diagnostic accuracy or impact were rated as using grade C or better methods (Table 7).

13.2 In spinal trauma and cervical myelopathy, studies show high concordance between abnormal neurologic status and abnormal MRI findings in the cord [180-187]. The incremental gain from using MRI for prognosis is unknown, and clinicopathologic correlations are not available. Most authors [181] suggest that CT is required for patients with suspected bony injury and that MRI can replace myelography.

13.3 Within the intervertebral disk, loss of magnetic resonance signal intensity indicates physiologic degeneration of the disk. The prevalence of signal loss increases with age in asymptomatic patients and is highest in the L4-5 and L5-S1 disks [188, 189]. The relation between loss of disk signal and back pain syndromes is not clear. In patients with lumbar herniated disks, shrinkage of the herniated fragment seen on MRI correlates with symptomatic recovery in patients who do not have surgery [190].

13.4 In the postoperative spine, MRI differentiates scar tissue from recurrent disk herniations better than does plain or postmyelogram CT [191, 192]. Paramagnetic contrast agents enhance scar tissue and may improve definition of scar from disk [193].

13.5 Diagnostic Accuracy

For causes of lumbar radiculopathy, a recent meta-analysis indicated that MRI, plain spinal CT, and myelographic CT had equivalent sensitivities and specificities, although myelography was less accurate [194, 195]. A carefully controlled prospective study in Wisconsin [196] and an older study [197] suggested the same equivalence for plain spinal CT, myelographic CT, and MRI. In another meta-analysis [198], MRI and CT had equivalent accuracy for spinal stenosis; both were better than myelography (Table 7). For patients with cervical radiculopathy, MRI and myelographic CT were highly concordant in showing nerve root narrowing or cord impingement [199-201].

13.6 Studies [202, 203] of MRI in asymptomatic persons suggest that between 10% and 30% have herniated or bulging disks (Table 7). Similarly, from 5% to 30% of several series of asymptomatic patients had spinal stenosis by imaging, which could be considered false-positive results, depending on how abnormalities were defined [198]. Incidental asymptomatic anatomic findings of disk herniation or bulge are equally as common (10% to 30%) in the thoracic and cervical spine as in the lumbar spine [204, 205]. Asymptomatic herniations appear in 20% of middle-aged persons and in more than 50% of persons older than 64 years [206].

13.7 Diagnostic Impact

In recent case series [39, 207-212] of spinal imaging, using MRI as the first test obviated the need for more than half of the CT scans and more than 90% of the myelograms. Although avoiding intrathecal injections with iodinated contrast is an obvious safety advantage, no studies rated as having good methods support the dramatic shift to MRI that is occurring in most metropolitan centers.

14. Miscellaneous Diagnoses

14.1 Case series involving patients with various diseases give information about the performance of MRI for a broad range of patients. Only a few early studies met minimal standards for methodologic quality, and all were based on 0.15-T machines [2]. Studies since 1987 have used higher field strength magnets with improved surface coils and paramagnetic contrast agents. Of three studies with various final diagnoses, only one met minimal standards (grade C) for methodologic quality [29]. The sensitivity of 1.5-T MRI was higher than the low-field MRI or standard CT (Table 3). The apparent specificity for 1.5-T MRI was lower, however, and the accompanying receiver-operator characteristic curves suggested equivalent overall accuracy. Other studies were flawed by lack of a defined reference standard for diagnosis [213] or by lack of control for test review or diagnosis review biases [11, 214].

14.2 In the brainstem and posterior fossa, MRI shows more lesions of all types than does CT [2, 215]. In a well-designed and informative study [216], Scottish investigators randomized the choice of first test for a large series of patients with suspected posterior fossa problems, then allowed clinicians to proceed as they wished with the remainder of patient care decisions. When MRI was the first test, 6% of patients also had CT, whereas 19% of patients had MRI when CT was the first test (Table 2). Magnetic resonance imaging changed 23% of diagnoses made by CT, whereas CT changed 18% of MRI diagnoses; however, this result was not statistically significant. The major reason given for getting the other test after CT or MRI alone was a desire to further clarify the diagnosis. One other study [217] reported that MRI changed the final diagnosis in 16% of patients.

14.3 Therapeutic Impact and Patient Outcomes

Several studies have reported changes in treatment plans after MRI (see Table 2). In the Scottish study [216] described above, MRI after CT changed 6% of treatment plans for patients with posterior fossa disease, and CT after MRI changed 2% of treatment plans. In an observational study, MRI resolved uncertainty about treatment choices in 76% of 113 patients and redirected the treatment choice in 39% of 69 patients [213]. This study was biased toward patients with previously unresolved diagnoses. Quality-of-life survey scores from the patients were unchanged from before the MRI up to 4 months after the MRI [213]. In a large prospective survey [218] from England, MRI findings altered treatment for 27% of patients. However, functional disability scores worsened, a finding attributed to the underlying serious illness present. Other reports of changes in treatment because of MRI used inadequate study designs (grade D) [11, 39, 119] or did not quantify findings [217].

15. Other Diseases and Syndromes

15.1 Cranial nerves can be imaged with high resolution MRI [219]. With MRI and magnetic resonance angiography, vascular compression of the preganglionic origins of the trigeminal nerve and the facial nerve has been associated with trigeminal neuralgia and hemifacial spasm, respectively [220-223].

15.2 In common clinical syndromes where the prevalence of structural disease is already low, the diagnostic yield of MRI also is low. In several studies of headache, no clinically significant mass lesions were found [224-226]. Among patients with dizziness, abnormal MRI findings were as prevalent in controls as in the dizzy patients and no treatable causes could be identified [227, 228].

15.3 Magnetic resonance imaging may define structural correlates of parkinsonian syndromes and other degenerative diseases of brainstem or cerebellar nuclei [229, 230]. Other morphologic abnormalities of the brain seen on MRI may correlate with mental illness, such as depression [231, 232], tardive dyskinesia [233], or primary schizophrenia [234]. However, clinical usefulness has not been shown.

16. Discussion

16.1 Two conclusions emerge from this review of the clinical efficacy of MRI for neuroimaging. First, MRI yields high-quality images. Indeed, most of the published studies show the visually impressive, technical capabilities of MRI as a new imaging technology. Second, although MRI holds great promise for improving clinical management, quantitative rigorous assessments of the clinical effect of MRI in large case series or well-controlled comparison trials are missing. Only a few studies document rigorous attempts at fair comparisons with other comparable technologies [156, 196, 216].

16.2 Formal assessment of the clinical efficacy for MRI has lagged behind its diffusion into clinical use for neurologic problems. Despite the installation of more than 2000 magnetic resonance scanners and the appearance of more than 5000 citations about neuroimaging with MRI, fewer than 30 studies are prospective controlled comparisons of diagnostic accuracy or changes in therapeutic choices, and no study documents a change in patient outcomes. Although the number of better-quality studies has increased from 6 in 1986 [2] to 29, this is fewer than 0.5% of the total of published reports about MRI. Editors of journals and reviewers of study proposals need to encourage more accurate and realistic appraisals of this valuable technology.

16.3 The way to improve diagnostic accuracy studies is already clear because the basic principles for completion of such studies have been known for more than a decade [235]. Because surgical confirmation of imaging findings often is impossible, determination of the final diagnosis may require several months of clinical follow-up and review of the total case by a "gold standards committee" [156, 196]. For assessment of diagnostic and therapeutic impact, the most successful protocol is the "randomized choice of first test" [216]. This study design is one of the few that can be followed through to an assessment of long-term patient outcomes.

16.4 The lack of high-quality clinical research and the widespread use of MRI pose a dilemma for evidence-based clinical technology assessment. The best compromise—between uncritical acceptance of usual clinical practices and silence about the appropriate guidelines for lack of evidence—is to present a set of suggested approaches, along with information about the quality of research used to support the suggestions. The Position Paper in this issue [4] gives such suggestions.

16.5 The trade-off between a higher sensitivity and a lower specificity (higher false-positive rate) for MRI has not been explored fully. A basic principle of signal detection theory and diagnostic testing is that more sensitive tests also have higher potential false-positive rates [236], and MRI is no exception. Because MRI is more likely than CT to show incidental findings or inconsequential pathologic results, MRI has a higher potential for false-positive interpretations than does CT. Determining the ultimate effect of such potential errors requires analysis of the consequences of false-positive and false-negative imaging results.

16.6 When MRI is equivalent to other diagnostic technologies, higher costs are a barrier to recommendations for MRI as the leading diagnostic test for neuroimaging. Costs can be decreased by finding more time-efficient imaging protocols, by decreasing the capital costs of scanners, and by restricting scanning to those patients most likely to benefit. Time efficiency may be achieved by development of faster imaging techniques and better receiver coils or by selective imaging focused only on the most important diagnostic possibilities. Either way, shorter imaging times would greatly improve the operating efficiency of MRI centers, which in turn should lower per-patient costs.