Hackmack K, Weygandt M, Wuerfel J, Pfueller CF, Bellmann-Strobl J, Paul F, Haynes JD. Can we overcome the 'clinico-radiological paradox' in multiple sclerosis?J Neurol. 2012 Mar 24. [Epub ahead of print]
The association between common neuroradiological markers of multiple sclerosis (MS) and clinical disability is weak, a phenomenon known as the clinico-radiological paradox. Here, we investigated to which degree it is possible to predict individual disease profiles from conventional magnetic resonance imaging (MRI) using multivariate analysis algorithms. Specifically, we conducted cross-validated canonical correlation analyses to investigate the predictive information contained in conventional MRI data of 40 MS patients for the following clinical parameters: disease duration, motor disability (9-Hole Peg Test, Timed 25-Foot Walk Test), cognitive dysfunction (Paced Auditory Serial Addition Test), and the expanded disability status scale (EDSS). It turned out that the information in the spatial patterning of MRI data predicted the clinical scores with correlations of up to 0.80 (p < 10(-9)). Maximal predictive information for disease duration was identified in the precuneus and somatosensory cortex. Areas in the precuneus and precentral gyrus were maximally informative for motor disability. Cognitive dysfunction could best be predicted using data from the angular gyrus and superior parietal lobe. For EDSS, the inferior frontal gyrus was maximally informative. In conclusion, conventional MRI is highly predictive of clinical disability in MS when pattern-based algorithms are used for prediction. Thus, the so-called clinico-radiological paradox is not apparent when using suitable analysis techniques.
For many years there has been a disconnect between what the MRIers are seeing and what is happening neurologically to the MSers, for example beta interferon can be great at quelling some of the MRI lesions but this was not reflected by such marked clinical benefit in movement. This may have been not that surprising because the clinical feature is a product of effects within the spinal cord and the brain and so without taking the spinal cord into account, MRI of the brain may not be that great in predicting what occurs.
In addition MRI suffered and suffers from poor resolution in that it was not really sensitive enough to really tell us enough. Then there is the fantasy effect, whereby MRIers use the imaging of MS to ascribe the role of some imaging sequence to a pathological effect, without experimentally proving that their imaging events is actually detecting what is said to be detected. Do we really know what a T1 or T2 lesion is?, is gadolinium measuring inflammation? when it is more likely to be measuring fluid rather than cellular movement. As we have said before an r value of correlation of 0.5 or 0.6 is not really a correlation when a really good correlation is 1 (direct relation ship more of one means more of the other) or -1 (indirect relationship more of one means less of the other) and no correlation is near 0. In this study they get correlations up to 0.8 (not bad) and what they say if you look in certain areas of the brain, rather than globally and look every where, then you can better idea of what symptoms may occur. This is logical because damage to certain nerve circuits lead to certain symptoms so if you look in these areas it is not surprising that you can link the load of the lesions to certain symptoms.