The February 2007 issue of the American Journal of Neuroradiology reported the findings of A. Ciafoni et al.1 In this article, the authors reported the approach of examining glutamate/glutamine peaks in 1H-MR (proton) spectroscopy to increase the specificity in differentiating tumefactive demyelinating processes from neoplasms. The thought of using MR spectroscopy in this situation is attractive because the current clinical approach may warrant a lumbar puncture and even biopsy. Although we commend the authors’ enthusiasm for this technique, the paper lacks statistical significance and contains much bias to support the authors’ claim that MR spectroscopy “can be helpful in the noninvasive diagnosis of acute demyelinating diseases”1 when elevated glutamate/glutamine peaks are identified. We believe that this article should be represented as an idea backed by case reports and not be considered as original research.
First, the authors never fully give the inclusion or exclusion criteria for how the 4 patients in the study were chosen. Were any patients excluded from the study, and, if so, why? They state that the spectroscopy was performed on 4 patients with neurologic symptoms and a diagnosis of parenchymal mass lesions on MR imaging. It seems unlikely that the first 4 patients who entered the study had the diagnosis of tumefactive multiple sclerosis. There must have been some patients originally selected who ended up with the diagnosis of a neoplasm. Why were these spectroscopy results not included in the study? They fail to notify the reader if the diagnosis was already known at the time of selection and if the interpreter of the spectral data was blinded. Obviously, these effects can cause bias, which would alter the outcome of the study. The potential for bias is evident if one examines the 2.1 to 2.5 region of the spectra where complex, unresolved resonances produce a “shoulder” of variable slope downfield from the N-acetylaspartate resonance. The fluctuations in the spectral intensity in this region are similar to the background noise (eg, Fig 2), which makes the assignment of a single peak height for the entire range challenging. Unfortunately, the authors have not described in detail their method for determining β, γ-Glx peak heights, and they have not documented the accuracy of the method for detecting changes in glutamine/glutamate levels (eg, in phantoms).
Second, no control is used. Referencing the article by Majos et al2 is not a substitute for using a control group with known neoplasms. Showing that a neoplasm does not increase the glutamate/glutamine peak would have added significantly to this article. Also, other demyelinating diseases are not included.
Finally, no statistical inference can be made in a study with only 4 patients; the study lacks statistical power.
For these reasons, we believe that this study should not be represented as original research, but as a novel idea with associated case reports to back up the need for more comprehensive research in this area. The facts in this study are not significant enough to suggest that MR spectroscopy “can be helpful” in distinguishing between tumefactive demyelinating diseases and neoplasm.
- Copyright © American Society of Neuroradiology