Valid measures of periodic leg movements (PLMs) during a suggested immobilization test using the PAM-RL leg activity monitors require adjusting detection parameters for noise and signal in each recording

Abstract

Objective and background: Individuals with restless legs syndrome (RLS) (Willis-Ekbom disease [WED]) usually have periodic leg movements (PLMs). The suggested immobilization test (SIT) measures sensory and motor features of WED during wakefulness. Surface electromyogram (EMG) recordings of the anterior tibialis (AT) are used as the standard for counting PLMs. However, due to several limitations, leg activity meters such as the PAM-RL were advanced as a potential substitute. In our study, we assessed the validity of the measurements of PLM during wakefulness (PLMW) in the SIT for PAM-RL using both default and custom detection threshold parameters compared to AT EMG.

Methods: 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's 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.

Results: 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=-13: -51.2, 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; -20, 10) and no significant magnitude bias. Two scorers independently scoring 13% of the SITs showed an adequate interscorer reliability of 0.96-0.98.

Conclusions: 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.

Keywords: Leg activity monitors; PAM-RL; Periodic leg movements during wakefulness (PLMW); Polysomnography; Restless Legs Syndrome (RLS); Suggested immobilization test (SIT); Willis Ekbom disease (WED).

Fig. 1

Minute example showing periodic limb movements during wakefulness detection from electromyogram and, custom and default PAM-RL.

Fig. 2

From the top: electromyogram (EMG) recording, custom PAM-RL, default PAM-RL RL (cPAM-RL) vs polysomnography (EMG) periodic limb movements during wakefulness with best fit regression with 95% confidence interval range. Max PAM-RL indicates maximum of Right or left leg was used for the PAM-RL values.

Fig. 3

Histograms showing the measurement errors for the PAM-RL compared to the periodic limb movements during wakefulness (PLMW) per hour defined by the electromyogram. (A) Default parameters for PAM-RL show marked skew and a significant error bias toward underestimating PLMW. Mean + standard deviation (SD) −31.7 ± 50.9, skew = −1.4, kurtosis = 2.1, median: quartiles = −13: −51.2, 0.0). (B) Custom parameters for PAM-RL show a more balanced distribution of errors with no significant skew and only a minimal error bias toward underestimating PLMW. Mean ± SD = −4.5 ±51.1, skew = −0.2, kurtosis = 2.6, median quartiles = 0.0 (−20, 10). Both distributions are leptokurtic with strong central tendencies and more extreme values than expected for a normal distribution. The median and quartile ranges therefore describe the distributions somewhat better than the mean and SD.

Fig. 4

Bland–Altman plot showing the distribution of differences between both (A) default and (B) custom PAM-RL periodic limb movements (PLMs) and electromyogram (EMG)-PLM (y-axis) against the average PLMs of PAM-RL and EMG (x-axis). There is clearly a big underestimation (mean, −31.7) by the default PAM-RL. The negative bias is mostly gone when PAM-RL onset and offset lines were customized. The mean bias of −4.5 indicated a slight underestimation by the cPAM-RL.

Fig. 5

Linear regression for suggested immobilization test periodic limb movements during wakefulness per hour from parameters independently set by RA and MY (r = 0.996).