Abstract Introduction Wearable sensors are increasingly used to collect personalized light exposure measurements in studies related to sleep and circadian rhythm, however little is known about their comparative validity. Recently, we showed that four commonly used wearable light sensors drastically differed in detectable range, accuracy, and variability for laboratory-measured photopic and melanopic illuminance. Here, we compared free-living light measurements for the two sensors with the most divergent lab-based performance. Methods As part of an ongoing clinical trial (NCT05398783), we measured free-living light exposure using Condor ActLumus and ActiGraph wGT3X-BT devices in 31 individuals (22F/9M; 36.5±15.9 yrs). Sensors were concurrently worn on the non-dominant wrist for a week, throughout the day, with randomized proximal and distal placement. The data was processed at 60s intervals, and non-wear time was detected using a modified Troiano’s method based on acceleration. Only periods of simultaneous wear on days with a minimum of 10h were analyzed. Results The mean daily light exposure per participant was strongly correlated between ActiGraph and Condor devices (R=0.91, p 0.001). However, the output of the ActiGraph was significantly less than that of the Condor (77.8 ± 67.5 vs 483.0 ± 456.0 lux, respectively, p 0.001). The largest discrepancy between the two devices was observed in their respective detection of light intensities below 100 lux. ActiGraph recorded significantly greater wear time per day spent at 0 lux than Condor (88.4 ± 6.9% vs 42.1±15.0%, p 0.001). On the contrary, ActiGraph recorded a significantly lower percentage of wear time in light intensities between 1 and 100 lux than Condor (4.7 ± 2.8% vs 39.9 ± 13.0%, p 0.001). Time spent in bands above 100 lux were also significantly greater for Condor (p 0.001), but discrepancies in wear time between devices were 10%. Conclusion The large observed differences in free-living light measurements were likely due to the inter-device discrepancies in detection limits and accuracy shown in our laboratory tests. These results demonstrate that assessing sensor performance is important to accurately interpret free-living light exposure measurements across studies using different sensors. We will further investigate how such differences may impact the determination of sleep timing and duration. Support (if any)
Ishihara et al. (Fri,) studied this question.