I read with interest the American College of Physicians' position paper on imaging of the brain and spine [1]. Sophisticated users know that MRI is a superior technique for detecting most diseases of the central nervous system. Neurologists and neurosurgeons routinely use MRI in their clinical practice because it gives them the most sensitive and specific indicator of clinically significant disease and has no deleterious effects. The alternative to MRI is a combination of many other tests, which, in the aggregate, can be more expensive. In patients with multiple sclerosis, for example, physicians might do invasive cerebrospinal fluid studies followed by brain stem evoked responses, visual evoked responses, and so forth. In patients with multiple sclerosis, one MRI study is more sensitive than all of these studies combined. It is important to note that MRI was used as the arbiter of drug therapy in the recent study of ß-seron in multiple sclerosis. Magnetic resonance imaging was acknowledged by neurologists as well as the Food and Drug Administration to be more sensitive than the clinical examination.

It would be taking a step backward to presume that the technique is suspect simply because studies do not fulfill the criteria proposed by the position paper. The authors have set up a strawman hypothesis in that they studied diagnostic accuracy and chose studies with more than 35 patients. There is no question that a study should include as many patients as possible; however, in studies of many neurologic diseases, it is difficult to enroll large numbers of patients. For example, it is difficult to get 100 patients with arteriovenous malformations, subarachnoid hemorrhage, aneurysms, white matter diseases of various kinds, and so forth without funding for multi-institutional trials. Who would fund these trials? Nobody would argue that the most powerful type of study is a randomized case–control study; however, the cost of these studies can be prohibitive.

The proof of the utility of MRI is definitely shown by the marketplace—a factor the authors do not address. The clinical efficacy of MRI is obvious because of its demand in neuroimaging. If MRI was not clinically useful, one would not have seen the extreme growth in the technology. In contrast, positron emission tomography was introduced in the mid-1970s but is still not widely used clinically to diagnose central nervous system disease.

What are the alternatives to MRI? All are slightly to extensively invasive. Computed tomography, for example, is more invasive, is not much less expensive than MRI, and is much less sensitive to diseases affecting the brain and the spinal cord.

We should ask not only about the positive predictive value of a study but also about what occurs when the MRI result is negative. Again, this has a significant effect on reassuring the patient and the practicing physician and is, in my opinion, as helpful as a positive result.

Kent and colleagues [2] are correct that MRI may not be the technology of choice to diagnose acute stroke; the diagnosis is made by looking at the patient directly. The imaging study is done to ensure that the patient is not suffering from a hemorrhage. Computed tomography is the initial imaging study of choice. If you plan to treat the patient with anticoagulant agents, it is important to know if the patient has bled.