The natural course of multiple sclerosis (MS) is slow and difficult to monitor clinically. A vital need exists, therefore, to seek paraclinical assessments such as MRI to monitor the disease accurately. Many new gadolinium-enhancing lesions—which reflect the inflammatory, active phase of demyelinating lesions—remain clinically silent, sometimes even when they involve otherwise eloquent regions, such as the optic nerve,1 brainstem and spinal cord.2 In fact, neuronal or axonal damage that leads to atrophy, which was previously considered to represent a late phenomenon in the secondary, progressive phase of the disease, has been revealed by MRI to occur relentlessly from disease onset. This damage might initially be clinically silent because of the redundancy of pathways in the CNS, and the capacity of the CNS for restoration and functional cortical adaptation.3
...neuronal or axonal damage ... has been revealed by MRI to occur relentlessly from disease onset
The ability of MRI to detect subclinical lesions has led to its widespread acceptance as a tool in the diagnostic work-up of patients who are suspected of having MS4 and in obtaining prognostic information early in the course of the disease.5 In addition, screening of new therapies to combat MS has been revolutionized by the use of MRI, which is more sensitive, more reproducible and has more favorable statistical properties than clinical measures, at least in the short term.
Many clinicians remain puzzled by the subclinical nature of MRI lesions in MS, and have sought to establish relationships with the traditional measures of outcome (that is, relapse rate and disability). Such an endeavor is evidently needed to understand how histopathological changes (as depicted by MRI) dictate clinical evolution in the long term, but this is a daunting task, as illustrated in a new paper by Daumer and colleagues.6 In this study, the authors question the use of MRI as an outcome measure in MS clinical trials. The multicenter database of the Sylvia Lawry Centre for MS Research was used by the authors to investigate whether baseline MRI findings would predict relapses and short-term disability measures in a pooled analysis of placebo-treated patients from a large number of completed trials. Baseline gadolinium-enhancing MRI lesions failed to predict the on-study relapse rate (whereas the prestudy relapse rate did), and baseline lesion load on T2-weighted MRI had minimal predictive value for subsequent changes in disability scores in the subgroup of patients with secondary progressive MS.
The results from the study need to be carefully considered, as the database used by Daumer and colleagues6 is unique in its complexity and hidden sources of bias, such as massive attrition due to incomplete data, heterogeneous definitions of clinical measures, variable MRI analysis techniques,7 and insufficient duration of follow-up. In fact, the same database has been used to demonstrate that the duration of follow-up in a typical randomized clinical trial is insufficient to determine disability changes with any confidence.8 The fundamentally negative findings of their study—failure to relate clinical findings of disease activity and progression to MRI markers—merely show that clinical measures are insensitive to the underlying pathological process, rather than indicating a shortcoming of MRI.
Clearly, validation of MRI findings is important, and much has been accomplished at the histopathological level. Gadolinium enhancement has been shown to reflect increased blood–brain barrier permeability, which is associated with inflammatory changes and phagocytosis of myelin by macrophages. Furthermore, in eloquent areas, such as the spinal cord and optic nerve, new enhancing MRI lesions are almost always associated with the occurrence of new clinical symptoms and signs.1, 2 However, for obvious reasons, many parts of the CNS are clinically difficult to assess. For example, lesions in the prefrontal cortex will not produce overt neurological symptoms or signs. Most new enhancing lesions leave a T2-hyperintense footprint after resolution of edema, but the histopathological importance of T2 lesions in general is heterogeneous; such lesions might even include remyelinated areas with presumably normal function. For this reason, and also because of the diverse locations of the T2 lesions, strong correlations between MRI findings and clinical disability are, in fact, very unlikely.9
Although the clinical effect of each and every new MRI lesion is unlikely to be determined with current clinical measures, patients would understandably prefer to have a scan devoid of new enhancing or T2 lesions. Likewise, following treatment, patients would prefer to have no new MRI lesions on their scans. This intuitive notion is reinforced by work that addressed the correspondence between treatment effects observed clinically and on MRI. Sormani et al.10 convincingly show in a meta-analysis of published, randomized trials that the effect sizes (if any) of experimental treatments for MS measured by MRI and by traditional clinical outcomes are indeed closely correlated; either they both show a similar positive effect, or they both show no effect. This study demonstrates that MRI fulfills the most important final step in validation of a surrogate marker according to the Prentice criteria. By contrast, the paper by Daumer et al.6 only analyzed placebo arms of the trials and, as a consequence, could not by definition meet such criteria.
Nevertheless, T2 lesions fail to depict the whole story of MS, and other MRI markers need to be defined and validated to understand the accumulation of disability in the long term. These markers include the determination of spinal cord pathology, general and regional cerebral atrophy, and damage to normal-appearing (on MRI) brain tissue, especially the gray matter.11 How these features can be integrated into a unified model remains to be established, and their inclusion should depend on the type of disability to be predicted. For example, spinal-cord MRI features are clearly needed to predict changes in currently used disability scales, whereas cerebral MRI changes are more suited to predict cognitive impairment.
In conclusion, MRI is, in our opinion, the desired outcome measure in MS clinical trials and the work of Ebers and colleagues,8 which highlights the weaknesses of the disability scale score to determine progression of MS in clinical terms, only adds further weight to this assertion. We hope that the negative findings by Daumer et al.6 will stimulate endeavors to better understand the relationship between clinical findings and (MRI depicted) histopathology, by developing improved clinical measures. Initial attempts to do so, such as the inclusion of cognitive testing in a functional composite score, have already indicated that the relationship between MRI and clinical findings can be improved,12 but will require very long periods of follow-up.5, 13
