Data were obtained from 39 participants who were diagnosed with primary WED and who were on stable medication as part of another study using the SIT to repeatedly evaluate WED symptoms over 6-12 months. EMG recordings and PAM-RL, when available, were used to detect PLMW for each SIT. Complete PAM-RL and polysomnography (PSG) EMG data were available for 253 SITs from that study. The default PAM-RL (dPAM-RL) detected leg movements based on manufacturer鈥檚 noise (resting) and signal (movement) amplitude criteria developed to accurately detect PLM during sleep (PLMS). The custom PAM-RL (cPAM-RL) similarly detected leg movements except the noise and movement detection parameters were adjusted to match the PAM-RL data for each SIT.
The distributions of the differences between either dPAM-RL or cPAM-RL and EMG PLMW were strongly leptokurtic (Kurtosis >2) with many small differences and a few unusually large differences. These distributions are better described by median and quartile ranges than mean and standard deviation. Despite an adequate correlation (r = 0.66) between the dPAM-RL and EMG recordings, the dPAM-RL on average significantly underscored the number of PLMW (median: quartiles = 鈭?3: 鈭?1.2, 0.0) and on Bland-Altman plots had a significant magnitude bias with greater underscoring for larger average PLMW/h. There also was an adequate correlation (r = 0.70) between cPAM-RL and EMG but with minimal underscoring of PLMW (median quartiles = 0.0; 鈭?0, 10) and no significant magnitude bias. Two scorers independently scoring 13% of the SITs showed an adequate interscorer reliability of 0.96-0.98.
Our study confirms our expectation that measuring PLMW in a SIT using dPAM-RL is not valid and that adjustments to the detection threshold criteria are required. The PAM-RL, using parameters customized for each SIT provided a valid and reliable measure of PLMW with minimal magnitude bias compared to the AT EMG recordings.