Improving test-retest variability of visual-evoked responses in multiple sclerosis: implications for trial design.

Remyelination is an important repair strategy in multiple sclerosis. Latencies of visual-evoked responses are a suitable surrogate for remyelination of the optic nerve. Their test-retest variability has been incompletely evaluated, especially in pathologically delayed potentials. Visual-evoked potential was recorded twice, 2.1 +/- 3.1 (mean +/- SD) days apart, in 39 patients with definite or evaluated for multiple sclerosis. Acute optic neuritis and current steroid treatment were exclusion criteria. Mean and difference of the two recordings were calculated for latencies and amplitude, both before and after verification of cursor positioning by a physician blinded for the sequence of recordings. Before verification, the difference between first and second visual-evoked potential was -2.07 +/- 9.07 milliseconds for N75 latency, -1.18 +/- 8.02 milliseconds for P100 latency, and -0.06 +/- 2.71 muV for N75/P100 amplitude (n = 77 eyes, mean +/- SD). Independent verification judged two eyes as unsuitable for analysis. The differences in the remaining 75 eyes were reduced to -1.22 +/- 6.86 milliseconds (N75), -0.7 +/- 3.85 milliseconds (P100) and -0.04 +/- 2.53 microV (amplitude). These effects do not differ between delayed and nondelayed eyes. Similar to magnetic resonance imaging, use of evoked potentials in multiple sclerosis remyelination trials will require independent verification, ideally by a central evaluating facility. Reproducibility should be verified individually at screening.