Introduction
Until magnetic resonance imaging (MRI) entered clinical routine, evoked potentials had been the most important paraclinical tool to objectively detect pathological changes in multiple sclerosis (MS) [
1]. Ever since, evoked potentials have remained part of the routine diagnostic work-up in many centers. Yet current diagnostic criteria attribute a role only to visual evoked potentials to lend objective paraclinical evidence in a patient reporting a previous episode of visual impairment whereas the possible contribution of other evoked potential investigations is recommended to be further explored [
2]. Somatosensory evoked potentials (SSEP) cover the whole spinocortical pathway; they are robust and easy to perform. According to some authors, SSEP are among the most valuable electrophysiological tests in MS [
3] with the highest sensitivity [
4] and of prognostic value [
5‐
9]. However, with the broader availability of MRI including spinal cord (SC) MRI, the question has arisen whether SSEP are of value beyond MRI and how the findings of both methods relate to each other. To the best of our knowledge, only very few studies reported findings from both methods [
10]. In this study, we retrospectively analyzed disability, SSEP, brain MRI, and whole SC MRI with full axial coverage in a larger cohort of patients with early MS.
Discussion
To the best of our knowledge, this is the first study relating SC MRI with full axial coverage to SSEP data in early MS. In our cohort, SC MRI was more sensitive to SC involvement whereas tSSEP were more closely related to disability. We will discuss methodological issues, the underpinning of abnormal SSEP, and the value of both diagnostic measures as biomarkers [
14]. We will also acknowledge limitations.
Our cohort was in a very early stage of MS with a median disease duration of one month and a maximum disease duration of 2 years. The significantly higher percentage of patients with abnormal SC MRI compared to tSSEP is well in accordance with the current diagnostic criteria of MS [
2] incorporating only SC lesions detected by SC MRI but not abnormal SSEP. Yet we did not expect such a clear difference from the literature. While the percentage of patients with SC lesions (68%) was well in the range reported in the literature [
15‐
18], the percentage of patients with abnormal tSSEP was low (22%). Other studies reported abnormal findings in > 80% [
4,
6,
19] but had included patients more severely affected and in later stages. Because of the lack of a commonly accepted standard to quantify abnormalities for both tSSEP and SC MRI, we dichotomized our results (normal vs. abnormal) to compare sensitivity and performed correlation analyses in subgroups. The largest, in part overlapping, groups were these with abnormal SC MRI, and with normal tSSEP; both allowing for correlation analyses as, by definition, lesion volume from SC MRI was > 0 and P40 latency measurable. We also performed subgroup analyses in a fully parallel manner to treat both parameters equally.
In early studies [
20,
21], the substrate of SSEP changes was termed ‘lesion’ [
1]. Yet it has remained unclear to what degree this is visible on conventional T2-weighted MRI. Of note, studies using visually evoked potentials demonstrated abnormal findings in the absence of overt acute inflammatory activity (i.e. a history of previous optic neuritis) suggesting the possibility of further mechanisms leading to abnormal EP [
22]. However, our results are largely compatible with the ‘lesion’ hypothesis. In 161 patients, we only observed a single case with abnormal tSSEP in the absence of a corresponding lesion on MRI along the spinocortical pathway; in this single case, P40 latency was only slightly outside the normal range so that it remains open whether this finding is related to MS at all. More interestingly, our results also provide evidence that tSSEP contain hidden information in the normal range as we here observed a remarkable correlation with disability (EDSS) even surviving correction for the existence of SC lesions. This association may be of value for correlation analyses in large groups. Nevertheless, we do not see a realistic way to leverage this information in clinical routine (i.e., at the individual level).
Regarding the demonstration of SC changes to aid the diagnosis of MS, our results are clear. SC MRI was more sensitive in detecting SC changes and is, hence, the more sensitive diagnostic biomarker—in this respect attributing a meaning to tSSEP only in case of an unavailable SC MRI. However, tSSEP were more closely associated with disability than SC MRI. In addition, our analyses of groups and subgroups suggested complementary information of both methods. These findings deserve recapitulation. tSSEP correlated with disability even in the normal range. When dichotomizing results, tSSEP but not SC MRI was able to discern more from less disabled patients. In the group of patients with SC lesions, lesion volume correlated with disability but, of note, the subgroup with abnormal tSSEP still showed more disability than the subgroup with normal tSSEP. A simple and straight forward explanation for these findings is that of a threshold effect with SC MRI having the lower threshold for being classified as abnormal. Compatible with our findings, this should lead to more patients being classified as abnormal by the methods with the lower threshold. As only one (small asymptomatic) lesion suffices to be classified as abnormal SC MRI, abnormal findings in this low-threshold test are more likely to go along with little or no disability than do abnormal findings in the high-threshold test (i.e., tSSEP in our case). Again, compatible with our findings, differences in disability between patients classified as abnormal and normal may be higher in the high-threshold method. However, even in statistical models comprising both methods, abnormal tSSEP were still (i.e., independently) related to more severe disability. The latter finding cannot merely be explained by a threshold effect. Differences in lesions may not be visible on MRI but have effects on SSEP measures such as eloquence of location (i.e., lesion volume within the spinocortical projection), and destructiveness (e.g., degree of demyelination, axonal loss, and remyelination). Our results are compatible with these mechanisms but unable to differentiate further. Moreover, it should be kept in mind that prognostication by measures such as MRI-based lesion volume or electrophysiological latencies likely results from autocorrelation, meaning that patients with higher values of a measure at baseline will also show a stronger increase of this measure in the later course. In conclusion, our results (of a closer relation of tSSEP to disability compared to SC MRI) are compatible with a potential value of tSSEP as a prognostic biomarker in complementation of SC MRI. Last but not least, currently available evidence on the prognostic value of evoked potentials seems more robust [
5‐
9] than that of lesions detected by SC MRI [
16‐
18,
23‐
25] although we are not aware of a study having directly compared both methods in this respect.
We acknowledge the limitations of our study. The number of patients is unlikely to have been large enough to cover the heterogeneity of mechanisms by which lesions can cause symptoms. Tibial SSEP cover SC pathways only incompletely and it is likely a matter of cohort size to find patients with SC lesions not changing tSSEP but causing severe clinical deficits. Moreover, our conclusions may not apply to later stages.
In summary, our results point to SC lesions, as visible on T2-weighed MRI, as the main driver of abnormal findings of SSEP in early MS. Accordingly, whole SC MRI is the more sensitive diagnostic biomarker than tSSEP. However, as changes in tSSEP were more closely related to disability, our data are, in principle, compatible with a potential role of tSSEP as a prognostic biomarker in the complementation of MRI, which however necessitates direct investigation.