Proposed modification to data analysis for statistical motor unit number estimate.

Abstract

Motor unit number estimation (MUNE) is an important electrophysiological technique for quantitative measurement of motor neuron loss. Although commonly used, there is no consensus concerning the optimal procedure for statistical MUNE, particularly regarding several operator-dependent variables. To assess the variables, we analyzed 500 sequential, submaximal compound muscle action potential (CMAP) responses at three or four stimulus intensities in 10 controls and 10 patients with amyotrophic lateral sclerosis (ALS). In both controls and ALS patients, we found that posttest filtering data based on 20% or 25% windows or 2, 2.5, or 3 SD excludes <5% of data. Windows of 10% or 15% excluded <5% of data in controls but not in ALS patients. Excluding data based on +/-2 SD, the coefficient of variation for final MUNE was 12% in controls and 6% in ALS patients. Group sizes of 30 or 50 and sample sizes of 300 to 500 sequential CMAP responses per run yielded the lowest coefficient of variation. We propose that statistical MUNE data should be analyzed based on excluding data >2 SD from the mean, because this is operator independent, includes the majority of data, effectively excludes clearly outlying data, such as fasciculations or movement artifact, and has a reasonable coefficient of variation.