Abstract Introduction Wearable and nearable sleep-monitoring devices are rapidly expanding in clinical and consumer settings, yet their accuracy relative to in-lab polysomnography (PSG) remains incompletely validated. This study evaluated the agreement, predictive value, and clinical performance of a novel photoplethysmography-based activity device (PAD) compared with the gold-standard PSG. Methods Fifty-nine nights of paired PAD and PSG recordings were analyzed. Measures from the PAD included subjective sleep quality score (SQS), total sleep time (TST 1), sleep latency (SL 1), heart rate (HR 1), movement minutes, heart rate variability (HRV), interruptions, and device-derived AHI. PSG outcomes included TST 2, SL 2, sleep efficiency (SE), wake after sleep onset (WASO), periodic limb movement index (PLMI), AHI 2, and HR 2. Paired t-tests compared device outputs. Pearson correlation coefficients assessed relationships between PAD and PSG metrics. Linear regressions were used to determine whether PAD-derived variables predicted PSG outcomes. Results PAD significantly overestimated sleep duration and onset latency. PAD-TST 339.1 ± 106.6 min vs PSG-TST 306.6 ± 107.5 min (t=2.45, p=0.017), PAD-SL (48.9±27.0 vs 21.3 ± 26.0 min; t=5.97, p 0.001). AHI showed no meaningful device difference (17.5 ± 18.8 vs 17.1 ± 19.1 events/hr; p=0.898), and heart rate was similar (64.4 ± 6.7 vs 66.1 ± 12.3 bpm; p=0.207). SQS strongly correlated with PSG-TST (r=0.56, p 0.0001), SE (r=0.53, p 0.0001), and inversely with WASO (r=−0.49, p=0.0001), while PAD interruptions correlated with PSG-WASO (r=0.48, p=0.0002) and PLMI (r=0.45, p=0.0004). Movement minutes and HRV were weakly or non-related to PSG outcomes (all p0.07). Regression models confirmed predictive utility: SQS independently predicted PSG-TST (β=2.73, p=0.0018) and PSG-SE (β=0.55, p=0.0016), while interruptions predicted WASO (β=10.28, p=0.0001) and PLMI (β=8.21, p=0.0007). The WASO model explained the greatest variance (adjusted R2=0.33), followed by SE (adj R2=0.20) and TST (adj R2=0.19). No PAD variable predicted PSG-AHI (adj R2 0.01). Conclusion PAD reliably reflects sleep continuity measures (TST, SE, WASO, PLMI) but overestimates sleep latency and duration and does not reflect respiratory severity. With strong correlation and predictive power for sleep fragmentation, PAD may serve as a clinically useful home-based monitoring tool but should not replace PSG in diagnosing sleep-disordered breathing. Support (if any)
Steelman et al. (Fri,) studied this question.