Pitfalls, fallacies and false positive rates

True and false positive rates of auditory and visual evoked potentials, as well as multi-modal evoked potentials, were assessed in groups of 30 multiple sclerosis patients and 19 healthy age-matched subjects. Sensitivity (true positive rate) generally proved to be high confirming published reports. False positive rates, which have not been extensively reported previously, proved to be unacceptably high. This held for two different abnormality thresholds (2 or 3 SD above the mean). The explanation for this proved to lie in the nature of an evoked potential test, which may be regarded as a combination of multiple independent components. The combination method, used to label the evoked potential test as normal or abnormal, governs the false positive rate. When the abnormality of at least one component is sufficient to label an evoked potential as abnormal, then the high false positive rate soars with the number of components. To increase the value of clinical evoked potential studies, it is advised that attention is paid to combination methodology, and that true as well as false positive rates should be directly measured.