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


Magnetic resonance imaging was better than computed tomography for diagnosing multiple sclerosis

ACP J Club. 1993 Nov-Dec;119:77. doi:10.7326/ACPJC-1993-119-3-077

Source Citation

Mushlin AI, Detsky AS, Phelps CE, et al. The accuracy of magnetic resonance imaging in patients with suspected multiple sclerosis. JAMA. 1993 June 23/30;269:3146-51.



To study the diagnostic accuracy of magnetic resonance imaging (MRI) in patients with suspected multiple sclerosis (MS).


A blinded comparison of MRI and computed tomography (CT) brain scans in patients with suspected MS.


Neurology clinics and office practices of neurologists and neuro-ophthalmologists in Canada and the United States.


303 consecutive patients (mean age 37 y, 222 women) referred because of suspected MS (possible or probable MS based on history and physical examination) who did not meet the Poser criteria for clinically definite MS. 190 patients (63%) were followed by a neurologist or other physician for at least 6 months.

Description of test and diagnostic standard

Axial dual proton density and T1- and T2- weighted images were done using a 1.5 tesla magnet scanner without intravenous contrast enhancement. CT scans used third-generation scanners and scans were done 1 hour after intravenous injection of 70 to 80 g of iodinated contrast material. Axial imaging was done using 5-mm-thick slices through the posterior cranial fossa and 10-mm-thick slices through the cerebrum. MRI and CT scans were assessed blindly by 2 neuroradiologists. Final diagnosis was decided by a panel of neurologists in a step-wise manner using clinical history and physical findings, laboratory tests including cerebrospinal analysis, evoked potential results, either CT or MRI findings, and then the other scan results (MRI or CT).

Main outcome measures

Receiver-operating characteristic (ROC) curve areas and likelihood ratios compared MRI and CT diagnosis with the final diagnosis, the latter excluding MRI and CT results.

Main results

20% of the patients had definite MS, 16% had probable MS, 16% had possible MS, 7% probably did not have MS, and 42% definitely did not have MS. The corresponding likelihood ratios for the MRI scan readings were 24.9, 2.9, 1.3, 0.3, and 0.3. Using the final diagnosis, the ROC curve area for MRI scans was 0.82 and was 0.52 for CT scans ( P< 0.001 for the comparison). Incorporating clinical information into the MRI scans improved the ROC area by 0.01 (to 0.83) ( P = 0.16). MRI and CT scans were equal in detecting non-MS abnormalities.


Magnetic resonance imaging was better than computed tomography for the diagnosis of multiple sclerosis. A "definite multiple sclerosis" reading after magnetic resonance imaging essentially established the diagnosis; however, normal imaging did not exclude the diagnosis of multiple sclerosis.

Sources of funding: Agency for Health Care Policy and Research; Blue Cross; Blue Shield; Berlex; General Electric Medical Systems.

For article reprint: Not available.


Magnetic resonance imaging for multiple sclerosis had high sensitivity and varying specificity

MS is a clinical diagnosis that is made best by a neurologist. Supportive tests include neuroimaging, evoked responses, urologic testing, and cerebrospinal fluid studies of IgG production and oligoclonal bands. These supportive tests help define research diagnostic criteria for clinical trials (1) and provide more diagnostic certainty for the patient. A potential diagnosis of MS causes great unease, prompting continued interest in confirming a diagnosis or ruling out MS as early as possible in patients with nonspecific neurologic symptoms.

AIS seen on MRI are found in MS and in a variety of other conditions including "healthy aging." The number, size, and location of AIS help differentiate MS from other neurologic diseases. The study by Offenbacher and colleagues adds important information on the sensitivity, specificity, and positive predictive value of several AIS diagnostic criteria. For an individual patient, the tables on positive predictive value and specificity will allow clinicians to assess just how supportive a given scan is of the clinical diagnosis of MS.

In a patient with 2 attacks of MS where clinical evidence of only 1 central nervous system lesion is present, MRI evidence of a separate lesion or positive test results for cerebrospinal fluid allow the diagnosis of definite MS to be made (1). These tests add to our knowledge about patients with monosymptomatic optic neuritis or acute myelopathy but do not allow definite diagnosis of MS to be made in such patients (1).

The study by Mushlin and colleagues confirms the utility of an MRI scan considered definite for MS (e.g., containing the multiple white matter and periventricular lesions). Unfortunately, only 36% of the patients with MS had such a characteristic MRI result. The low (1%) false-positive rate for such classic findings using MRI confirmed the diagnosis for approximately 20% of patients having an MS evaluation who were now spared the anguish of an equivocal diagnosis. Unfortunately, half of the patients sent for MRI in this study had a negative result (MRI report of "probably not" or "definitely not" MS), even though more than 25% of these patients were judged clinically to have MS. Thus, a negative MRI does not rule out MS. MRI scanning is the preferred imaging technique for MS. Mushlin and colleagues compared double-contrast CT imaging with MRI and found that MRI was markedly better. MRI is more likely to show abnormalities in a patient with MS than CT, evoked responses, or cerebrospinal fluid examinations (2). The false-negative rate of the combination of these studies is low but not sufficiently low to fully reassure the patient—watchful waiting is still necessary.

Progress in neuroimaging continues at a fast pace, and new techniques are being evaluated for patients with clinical syndromes prompting an MS evaluation (2). Additional studies will clarify the utility of paramagnetic contrast agents and alternative MRI techniques. Serial MRI scanning, with or without gadolinium-diethylenetriamine penta-acetic acid enhancement, has been used successfully to monitor treatment effects. Echo-planar MRI (allowing imaging after functional activation), magnetic resonance spectroscopy imaging (allowing quantitation of certain tissue components), positron emission tomography, and electroencephalographic imaging are all procedures that may improve our future diagnostic certainty for early or atypical MS, as will our abilities to provide patients with prognostic information and to better monitor their treatment.

Jay S. Luxenberg, MD
The Mount Zion Institute on AgingSan Francisco, California, USA


1. Poster CM, Pay DW, Scheinberg L, etal. New diagnostic criteria for multiple sclerosis: guidelines for research protocols. Ann Neurol. 1983;13:227-31.

2. Goodkin DE, Rudick RA, Ross JS. The use of brain magnetic resonance imaging in multiple sclerosis. Arch Neurol. 1994;51:505-16.