Reliability of EMG spectral parameters in repeated measurements of back muscle fatigue.

The change in median frequency of the power spectrum of the electromyographic (EMG) signal may be used as a measure of muscle fatigue. The reliability of the median frequency parameters was investigated for EMG-recording sites at L1 and L5 right and left on the erector spinae. The reliability of subjective fatigue ratings of the back muscles (Borg CR-10 scale) and of maximal trunk extension torque (MVC) was also investigated. Eleven subjects with healthy backs performed a 45-s isometric trunk extension at 80% of MVC twice a day, on three different days. Two-factor analysis of variance was made to obtain the different variances from which the standard error of measurement (SEM) and the intra class correlation coefficient (ICC) were calculated. The SEM within-day was somewhat lower than that between-days. Both were about the same at all four electrode sites. The 95% confidence interval for the studied variables was for the initial median frequency +/- 10 Hz, for the slope +/- 0.4-0.5%/s, for the MVC +/- 36 Nm and for the Borg ratings +/- 1.6. We conclude that, with the presently used method, changes or differences within these limits should be regarded as normal variability. The slope may be of limited value because of its large variability. Whether the low intraclass correlation coefficient for the EMG parameters in the presently studied test group implies a low potential in discriminating subjects with back pain can not be decisively concluded.