Current issues of ACP Journal Club are published in Annals of Internal Medicine


Review: More research is needed to evaluate clinical efficacy of MRI

ACP J Club. 1994 Sept-Oct;121:49. doi:10.7326/ACPJC-1994-121-2-049

Source Citation

Kent DL, Haynor DR, Longstreth WT Jr, Larson EB. The clinical efficacy of magnetic resonance imaging in neuroimaging. Ann Intern Med. 1994 May 15;120:856-71.



To evaluate the clinical efficacy of magnetic resonance imaging (MRI) for neuroimaging and to suggest guidelines for clinical practice.

Data sources

Relevant citations were identified by searching computerized databases, including MEDLINE, from 1987 through November 1993.

Study selection

Studies on the diagnostic accuracy of MRI alone or compared with other tests that had any effect on therapeutic decisions or patient outcomes were selected for review by independent readers. Reviews, technical reports, case reports, and studies describing < 30 patients were excluded.

Data extraction

Data were extracted on the disease studied, the level of clinical efficacy, and the methodologic quality of the study. Estimates of sensitivity and specificity were derived from studies with superior methodologic quality.

Main results

Of 3125 citations retrieved, 156 studies were selected and their methodologic quality was rated on a scale of A through D. Of the 143 rated for diagnostic accuracy or impact, 1 article about diagnostic accuracy had grade A quality, 28 were graded B or C, and 113 were graded D. 1 randomized trial and 2 comparison studies had grade B or C information about the effect on therapeutic choices. Safety evaluations indicated that noncontrast MRI is safe for most patients but costs of MRI are 30% to 100% higher than for computed tomography (CT). MRI shows more detail than CT and thus identifies more clinically silent abnormalities. For most abnormalities, the sensitivity of MRI is equal to or better than that of competing technologies. For example, the sensitivity of MRI for stroke and transient ischemic attacks ranges from 86% to 95%; for angiographically significant carotid stenoses, from 86% to 100%; and for multiple sclerosis, from 60% to 80%. MRI is more accurate for lesions in the temporal lobes, posterior fossa, brain stem, and spinal cord. For lumbar radiculopathy, MRI and plain spinal CT are as accurate as post-myelographic CT and are less invasive. CT is sufficient for initial diagnosis of most mass lesions or intracranial hemorrhages requiring immediate intervention. The role of MRI for carotid artery stenosis is being evaluated. A few studies found a modest effect of MRI results on therapeutic decisions but found no effect on patient quality-of-life or disability.


Magnetic resonance imaging yields high-quality images. The supporting evidence for appropriate use of magnetic resonance imaging in clinical practice in published studies is weak. Firm guidelines should be based on additional, more rigorous clinical research.

Sources of funding: In part, Department of Veterans Affairs Medical Center, Seattle, Washington; American College of Physicians; Agency of Health Care Policy and Research.

For article reprint: Ms. L.J. White, Director, Department of Scientific Policy, American College of Physicians, Independence Mall West, Sixth Street at Race, Philadelphia, PA 19106-1572, USA. FAX 215-351-2845.


These meta-analyses on MRI serve an important function. In addition to collecting and grading all relevant studies, the authors show how scant the data are on the utility of this important but expensive new technology. Little argument exists that the images obtained from MRI are anatomically more revealing than images using older technologies, such as CT. This improved resolution can reveal incidental findings and white matter changes that are difficult to interpret. For many neurologic conditions, however, it is unclear whether MRI adds clinically useful information commensurate with the extra cost.

MRI differs from CT in more than just resolution. All slices for a particular MRI study are taken simultaneously, whereas CT slices are imaged sequentially. In conditions such as dementias, a patient may be more likely to have a successful CT than an MRI because if the patient moves during the study, individual slices can be repeated. The MRI scanner's gantry is much more confining than a CT scanner. Patients with claustrophobia are more likely to tolerate CT. A clinician must be capable of individualizing the choice of imaging modality for the characteristics of each patient.

The guidelines published in conjunction with this paper (1) are a very useful summary of current knowledge, and are annotated with an indication of the relative strength of the evidence from which the recommendations were made. Because of the paucity of high-quality data on the relative benefit of MRI compared with alternatives, these guidelines should be regularly reevaluated and updated. Given the huge investment in MRI, a high priority should be assigned to the funding of the necessary comparative studies.

Jay Luxenberg, MD
University of California at San Francisco/Mount Zion Center on AgingSan Francisco, California, USA


1. American College of Physicians. Magnetic resonance imaging of the brain and spine: a revised statement. Ann Intern Med. 1994;120:872-5.