What is the validity of an “abnormal” evoked or event-related potential in MS?: Auditory and visual evoked and event-related potentials in multiple sclerosis patients and normal subjects

The predictive validity of evoked potentials (EPs) and event-related potentials (ERPs) in multiple sclerosis (MS) has not yet been fully investigated, as only the sensitivity of these tests has so far been reported. EPs (short, middle and long latency auditory evoked potentials and visual evoked potentials) and ERPs (visual and auditory) were studied in 19 controls and 30 multiple sclerosis (MS) patients. Abnormality thresholds of peak latencies were defined on the basis of the mean plus 2 or 3 standard deviations, based on data from the control group. The effects of changing the latency thresholds and including the absence of peaks in the abnormality definition were assessed. In accordance with earlier reports we found a high sensitivity (up to 93% for bimodal combined EPs and 47% for combined ERPs). False positive rates of separate peaks were low and conformed to expectation. However, combining separate peak measurements increased false positive rates of EPs and ERPs to unacceptably high levels (up to 58% for combined EPs and 32% for combined ERPs). Positive likelihood ratios for bimodal EPs were low (between 1.6 and 4.0, depending on the abnormality definition). They ranged from 1.4 to 2.2 for bimodal ERPs. Abnormal combined EPs or ERPs were therefore not the reliable indicators of functional damage that they are supposed to be. Separate EPs were much more reliable in this respect. ERPs failed to distinguish between the groups, either separately or in combination. Changing the latency threshold and including absent peaks in the abnormality definition influenced the abnormality rates in both groups. However, these changes were not consistent, and occurred in the patient as well as in the control group. As a result the validity of the tests was not improved. The choice of abnormality definition and the method of data combination have a pronounced effect on the false positive rates of EPs and ERPs, which are unacceptably high with currently conventional methods.