News and Publications

Assessing Information on T2-Weighted Magnetic Resonance Images of Patients With Secondary Progressive Multiple Sclerosis

J.A. Koziol, S. Wagner,* H.-P. Adams**

* University of Heidelberg, Heidelberg, Germany
** Humboldt University at Berlin, Berlin, Germany

Cranial magnetic resonance imaging (MRI) is a powerful procedure for diagnosing multiple sclerosis, delineating the natural history of the disease, and, potentially, assessing the response of multiple sclerosis patients to experimental therapy. The images clearly reveal the multiple, primarily periventricular, lesions that increase and decrease in size at different rates in separate regions of the brain. MR images show enhancement of multiple sclerosis lesions after injection of contrast medium because an abnormality in the blood-brain barrier produced by a perivenous inflammatory process allows the injected medium to cross the barrier. Indeed, contrast enhancement on MR images is widely accepted as a measure of disease activity in multiple sclerosis.

Nevertheless, disease-related activity, as measured by MRI, is a complex issue. The appearance of new lesions, gadolinium enhancement, enlarging lesions, and changes in the size of lesions on T2-weighted images have all been cited as potential measures of disease activity. The clinical utility of MRI in the assessment of multiple sclerosis patients is predicated on the assumptions that tissue types can be accurately and precisely classified on the basis of findings on MR images and that a relationship exists between the MRI findings and the clinical course of the disease. However, the reproducibility of MR images has inherent problems, many of which might be mitigated by appropriate adherence to standard protocols and appropriate preprocessing and postprocessing of the images.

Similarly, the typically weak correlations between the parameters used to determine clinical efficacy and MRI findings in multiple sclerosis may be partly attributable to methodologic limitations of cross-sectional studies with small, heterogeneous samples or of longitudinal studies with short follow-up periods. Because of the considerable interpatient variability in the distribution of lesions, an overall relationship between clinical activity and MRI findings might emerge only when a large cohort is studied over a long period.

We investigated the relationship between MRI findings and measures of clinical outcome in the context of a randomized, double-blind, placebo-controlled clinical trial to evaluate the efficacy of cladribine in the treatment of chronic progressive multiple sclerosis. Fifty-one patients were entered into the study. All 51 had had secondary progressive multiple sclerosis for more than 2 years, as shown by definitive clinical or laboratory findings. The clinical outcome of treatment in the trial was determined by assessments with 2 neurologic impairment scales: the Kurtzke Extended Disability Status Scale and the Scripps Neurologic Rating Scale. During the course of the trial, monthly scores on both scales were determined for each patient by the same neurologist, who had no knowledge of the patient's treatment group.

In accordance with the protocol for this study, patients had MRI at baseline (time of entry into the trial) and then at 6-month intervals during the 2 years of the trial. Two raters evaluated the baseline, 6-month, and 12-month T2-weighted images. One rater used a standard semiautomated quantitative technique; the other rater, a neurologist, did an independent evaluation of the images. The end point was comparison of the clinical outcome measures with the MRI rankings as determined by the 2 raters, with a statistical assessment of the degree of association. In particular, we wished to assess formally the statistical hypothesis that neither the quantification nor the neurologist's evaluation is strongly associated with the clinical course of disease.

The mean Spearman rank correlation between the 2 raters was .097 (standard error, .068) for the original rankings determined by the semiautomated quantitative technique and .126 (standard error, .066) for the rankings determined by evaluation of the images by the neurologist. Rank correlations between the raters and the clinical outcome measures were all less than 0.20 in absolute magnitude. We consider the reliability of both investigators high, yet the degrees of association between them, and with the clinical outcome measures, are all rather low.

Limitations with using T2-weighted images to determine the numbers and sizes of lesions may have contributed to the lack of association. More importantly, factors other than lesion load contribute substantially to the clinical assessment of the severity of multiple sclerosis. In particular, the lack of pathologic specificity of findings on conventional T2-weighted images may explain the lack of a strong correlation. Edema, inflammation, gliosis, demyelination, and axonal loss all result in increased signal intensity on T2-weighted images, but the effects of these changes on neurologic function most likely are different. Or, more than 1 pathophysiologic process may lead to demyelination in multiple sclerosis, and the sensitivity of T2-weighted images for the detection of each process is different. An immediate implication from this study is that more than quantitation of lesion sizes can be gleaned from T2-weighted MR images.

PUBLICATIONS

Koziol, J.A., Wagner, S., Adams, H.-P. Assessing information in T2-weighted MRI scans from secondary progressive MS patients. Neurology 51:228, 1998.