Emerald Touchless Sensor Quantifies Nighttime Scratching and Sleep Quality in Children With Atopic Dermatitis: PEDISTAD EMERALD Study

Why carry out this study? In pediatric populations with atopic dermatitis, itch-related nighttime scratching is associated with significant sleep disturbances. Characterizing the impact of pruritus and scratching on sleep patterns is a key unmet need in the pediatric population The objective of this study was to assess the feasibility of using the Emerald touchless sensor to quantify nighttime scratching and sleep patterns in pediatric patients aged 6 to 11 years with atopic dermatitis What was learned from this study? Results of the Emerald touchless sensor showed a strong correlation between increased scratching and poor sleep efficiency. Additionally, there were suggestive positive correlations between scratching metrics and patient-reported nighttime itch that did not reach statistical significance in this exploratory cohort of pediatric patients with atopic dermatitis It demonstrates feasibility in quantifying nighttime scratching and sleep quality in pediatric cohorts, which warrants investigation in larger, controlled studies

PEDISTAD EMERALD ( NCT05235724 ), a substudy of the PEDISTAD registry ( NCT03687359 ), was an open-label, 12-week, exploratory study conducted at US-based sites. The study included 20 pediatric patients aged 6–11 years who were originally identified with moderate-to-severe AD from the PEDISTAD registry. The study consisted of a screening period (day − 28 to − 14), the EMERALD initiation period (day − 14 to − 1), and a 12-week treatment period. During the treatment period, patients received SoC treatment for AD at the discretion of the investigator, including topical (e.g., topical corticosteroids, topical calcineurin inhibitors, and emollients) and systemic treatments (e.g., methotrexate, cyclosporine, systemic corticosteroids, dupilumab, and other immunological agents) (Fig. S1 in the electronic supplementary material [ESM]). Inclusion criteria were participation in the PEDISTAD registry at one of the US sites, age of 6–11 years at the time of enrollment, and Peak Pruritus Numeric Rating Scale (PP-NRS) score ≥ 4 at screening. Patients were required to have access to reliable and operational WiFi at screening and throughout the study period and to own a device (mobile phone, tablet, or computer) to record daily itch and sleep disturbances at screening and throughout the study period. They were required to sleep by themselves at least 5 nights a week for a minimum of 5 h a night, and signed informed consent by the parent/legally acceptable representative and assent by the participant appropriate to the patient’s age. The study was conducted in accordance with the Declaration of Helsinki, the International Conference on Harmonisation Good Clinical Practice guideline, and applicable regulatory requirements. An independent data and safety monitoring committee conducted blinded monitoring of patient safety data. The local institutional review board or ethics committee at each study center oversaw trial conduct and documentation. All patients, or their parents/guardians, provided written informed consent before participating in the trial. Pediatric patients provided assent in accordance with the Ethics Committee (Institutional Review Board [IRB]/Independent Ethics Committee)-approved standard practice for pediatric patients at each participating center. The study was reviewed and approved by Western Institutional Review Board (IRB approval no. 20216456). A complete list of all approving ethics committees is provided in Table S1 in the ESM. The Emerald touchless sensor (Emerald Innovations, Cambridge, MA) was provided to patients to quantify nighttime scratching and sleep quality and was mounted to the wall. Figure S2 presents photographs of the Emerald touchless sensor, showing its physical appearance (Fig. S2a in the ESM) and its placement in a home bedroom setting (Fig. S2b in the ESM). The Emerald sensor has a machine learning model that can identify scratching and differentiate it from other movements by leveraging the fact the fact that scratching is a repetitive, localized motion. The model has been programmed using a large, labeled data set in which patients with AD and chronic pruritus of varying severity were monitored simultaneously with the Emerald touchless sensor and a video camera. Trained human labelers then labeled scratching events in the video data, and a ground truth corpus of scratching events was created via a consensus procedure applied to the human labels. The machine learning model was trained and tested on this ground truth dataset using leave- n -out cross-validation, while ensuring that in each fold, patients were either in the training data set or the test data set. The key study endpoints are sleep efficiency (defined as the total time in minutes spent in a sleep stage [light, deep, or rapid eye movement; which were pooled for the analysis] as a percentage of the total time in bed) and sleep onset (or latency, defined as the time in minutes when the person enters the bed area and becomes relatively static, until the sleep start time, and feasibility of using the Emerald touchless sensor [measured as percentage of nights with evaluable data over the 12-week study period]). Please see Supplementary Materials in the ESM for full details of endpoints and definitions. The 20-patient sample size determination was based on operational feasibility rather than statistical considerations. Due to the exploratory nature of the study, statistical analyses consisted of descriptive analyses with no adjustment for multiple comparisons. Multiple correlation analyses were tested (sleep onset vs. PP-NRS; scratching events vs. PP-NRS; scratching time vs. PP-NRS; scratching events vs. sleep efficiency), thereby increasing the risk of Type I errors. The following correlations were examined: correlation analyses between weekly averages of Emerald-derived nightly scratching events per hour and Emerald-derived sleep efficiency over a 12-week period, correlations of Emerald scratching metrics (scratching events and time per hour) with patient- or caregiver-reported outcomes of itch as measured by PP-NRS, and correlation of sleep onset with patient- or caregiver-reported outcomes of itch as measured by PP-NRS (see Supplementary Materials available in the ESM for additional study outcomes). For these correlation analyses, Pearson or Spearman correlations were computed using the following categories: 0.8 to 1.0 (very strong), 0.6 to 0.8 (strong), 0.4 to 0.6 (moderate), and 0.2 to 0.4 (weak).

Among 18 patients enrolled in the PEDISTAD Emerald substudy after meeting the eligibility criteria at screening, 17 contributed data and were included in baseline characteristics (Table S2 in the ESM); sufficient study data were collected from 16 patients (2 patients were excluded because of noncompliance and insufficient data). Of the 16 patients included, the mean percentage of nights with evaluable data during the 12-week study period was 96.1% (standard deviation [SD]: 4.70%), and the median was 97.9% (range: 83.3–100.0%). The mean (SD) age of the study population ( n  = 17) was 7.2 (1.7) years and comprised 10 (58%) males and 7 (42%) females (Table S2 in the ESM). Correlation analyses between weekly averages of Emerald-derived time until sleep onset and patient- or caregiver-reported PP-NRS through week 12 were used to objectively assess the association between sleep onset and itch. Results show a positive correlation between PP-NRS and sleep onset (Pearson correlation coefficient: r = 0.50; p  = 0.049; Fig.  1 a), indicating the more itch a patient experiences, the longer the time to fall asleep. Fig. 1 a Correlation between sleep onset and PP-NRS (Pearson correlation: 0.50; p  = 0.049; 95% CI: − 0.02 to 0.81. b Correlation between Emerald sleep efficiency and Emerald scratching events per hour (Pearson correlation: − 0.66; p  = 0.003; 95% CI: − 0.87 to − 0.25) (Spearman correlation: − 0.66; p  = 0.004; 95% CI: − 0.87, − 0.24). c Correlation between Emerald scratching events per hour and PP-NRS; not statistically significant (Pearson correlation: 0.46; p  = 0.076; 95% CI: − 0.07 to 0.79). d Correlation between Emerald scratching time per hour and PP-NRS; not statistically significant (Pearson correlation: 0.43; p  = 0.096; 95% CI: − 0.10 to 0.77). CI confidence interval, PP-NRS Peak Pruritus Numeric Rating Scale a Correlation between sleep onset and PP-NRS (Pearson correlation: 0.50; p  = 0.049; 95% CI: − 0.02 to 0.81. b Correlation between Emerald sleep efficiency and Emerald scratching events per hour (Pearson correlation: − 0.66; p  = 0.003; 95% CI: − 0.87 to − 0.25) (Spearman correlation: − 0.66; p  = 0.004; 95% CI: − 0.87, − 0.24). c Correlation between Emerald scratching events per hour and PP-NRS; not statistically significant (Pearson correlation: 0.46; p  = 0.076; 95% CI: − 0.07 to 0.79). d Correlation between Emerald scratching time per hour and PP-NRS; not statistically significant (Pearson correlation: 0.43; p  = 0.096; 95% CI: − 0.10 to 0.77). CI confidence interval, PP-NRS Peak Pruritus Numeric Rating Scale Correlation analyses of weekly averages of Emerald-derived sleep efficiency and Emerald scratching events per hour through week 12 were used to objectively explore and quantify the association between nighttime scratching and sleep patterns. Data for the three sleep stages were pooled to obtain a single overall sleep efficiency estimate. Analyses revealed a strong negative correlation between scratching events per hour and median sleep efficiency (Pearson correlation coefficient: r = − 0.66; p  = 0.003) (Fig.  1 b), indicating that the more often patients scratch, the worse their sleep efficiency is. Lastly, correlation analyses between Emerald scratching metrics (events or scratching time per hour) and patient- or caregiver-reported PP-NRS through week 12 were used to objectively assess the association between nighttime scratching and itch. Results show a positive correlation between scratching events per hour (Pearson correlation coefficient: r = 0.46; p  = 0.076) and PP-NRS (Fig.  1 c) that did not reach statistical significance, suggesting that the worse a patients’ itch is, the more they scratch. Similarly, there was a positive correlation between scratching time per hour (Pearson correlation coefficient: r = 0.43; p  = 0.096) and PP-NRS (Fig.  1 d) that did not reach statistical significance. Figure S3 presents example data from a single patient, showing the progression across time of scratching events (Fig. S3a in the ESM) and scratching time (Fig. S3b in the ESM) from week − 2 through week 11 and illustrates how scratching events and time correlate with the patient-reported sleep numeric rating scale (sleep NRS; range 0–10) and PP-NRS.

This feasibility study shows that the Emerald touchless sensor can provide objective measures of itch-related scratching and sleep disturbance in children with AD, with high compliance in a home setting. While the strong correlation between scratching events and sleep efficiency is promising, the findings from this small exploratory cohort should be confirmed in future larger, controlled studies with diverse disease presentations, standardized treatments, and formal validation against gold-standard measurement tools.

The results from this study showed a high level of evaluable data, demonstrating high compliance and suggesting that the Emerald sensor has feasibility for use in pediatric clinical trial settings. This is particularly notable given that it was a study in a pediatric population over the summer, when many participants went on vacation, demonstrating high feasibility even outside controlled settings. In this small exploratory cohort ( n  = 16), this study also demonstrated a statistically significant correlation between increased scratching and poor sleep efficiency, in concordance with data from a similar study with Itch tracker, an Apple Watch-based app [ 29 ]. Other correlations between Emerald-derived scratching metrics and patient-reported PP-NRS approached but did not achieve statistical significance (scratching events vs. PP-NRS: p  = 0.076; scratching time vs. PP-NRS: p  = 0.096), suggesting preliminary feasibility of the sensor in this exploratory sample. The children included in this study were originally identified with moderate-to-severe AD at enrollment into the PEDISTAD registry. However, at baseline assessment for the PEDISTAD EMERALD substudy they demonstrated mild AD symptoms, as measured by the clinician-assessed outcome (Eczema Area and Severity Index of 0.1–5.9 is considered mild disease severity) [ 30 ] and mild patient- or caregiver-reported pruritus symptoms as measured by PP-NRS (PP-NRS of 0–3 is considered mild itch severity) [ 31 ], reflecting successful disease control on SoC treatment. Despite this well-controlled disease state, sleep disturbances persisted, highlighting the significant burden of AD [ 1 , 32 ]. Itch and sleep are often measured with patient- or caregiver-reported rating scales, which are subjective [ 33 ] and correlate only moderately with objective measures [ 16 , 19 ]. Objective assessment of nighttime itch and sleep patterns can provide a better understanding of AD disease burden and help guide effective treatment to minimize symptoms. The Emerald touchless sensor produces objectively measured data that can characterize itch-related scratching and sleep disturbance in patients with AD. Future studies are needed to assess whether the sensor can detect clinically meaningful changes that might not be captured by patient-reported outcomes. It has the advantage of continuously collecting physiological metrics in the home with high coverage and reliability, showing potential for long-term studies, even in pediatric populations. This study has several important limitations. First, the sample size is small ( n  = 16 analyzable patients), which limits statistical power and generalizability of findings. Given the small sample size, stratification or statistical adjustment for treatment type was not feasible. Also, multiple correlations were tested without correction for multiple comparisons, increasing the risk of Type I errors. Second, this was an open-label study without a healthy control group or comparison with other established measurement tools (e.g., video-derived scratching, actigraphy, polysomnography), which limits our ability to validate the Emerald sensor against gold standards in this pediatric cohort. Third, patients received heterogeneous SoC treatments, including topical therapies, systemic treatments (methotrexate, cyclosporine), and biological agents (dupilumab). This heterogeneity in treatment type and intensity represents a significant uncontrolled confounder that may substantially influence itch intensity, scratching frequency, and sleep patterns. Fourth, this study was conducted during summer months when many families were on vacation, which may have introduced seasonal variation in disease activity, sleep schedules, and environmental factors that could affect both scratching and sleep patterns. Fifth, inclusion criteria required reliable WiFi connectivity and single-occupancy sleeping arrangements (≥ 5 nights per week), which may have introduced selection bias and limits generalizability to families with different technology access or sleeping arrangements. Sixth, families were aware that the Emerald sensor was monitoring scratching and sleep, which may have altered scratching behavior, sleep patterns, or caregiving responses (Hawthorne effect). Given these limitations, these exploratory findings should be interpreted as preliminary and hypothesis-generating, requiring confirmation in future adequately powered studies with appropriate control groups, standardized or stratified treatments, and formal validation against established measurement tools.

Atopic dermatitis (AD) is commonly associated with debilitating itch (also known as pruritus), often resulting in scratching, which can profoundly impact sleep [ 1 , 2 ]. Sleep disturbances secondary to AD are associated with an overall reduction in health-related quality of life [ 3 ] and may be associated with a higher likelihood of attention-deficit hyperactivity disorder [ 4 ], daytime sleepiness [ 5 ], and increased sedentary behavior [ 6 ]. Furthermore, children with AD and poor sleep quality appear to be more susceptible to linear growth impairments [ 7 ]. Consistent with this, findings from the broader PEDISTAD cohort ( n  = 1329) revealed reduced height in male and female children aged < 12 years with moderate-to-severe AD compared with age norms [ 8 ]. Similarly, the phase 3 studies LIBERTY AD PEDS ( NCT03345914 ) and LIBERTY AD ADOL ( NCT03054428 ) reported that patients aged 6–17 years with moderate-to-severe AD were overrepresented in lower height percentiles [ 9 ]. Patient- and caregiver-reported surveys are used to assess sleep in pediatric patients; however, they are subjective [ 10 ]. In particular, the results may be biased when parents report on their children’s sleep and other causes of sleep disturbance [ 2 , 11 ]. Objectively characterizing the impact of pruritus and scratching on sleep patterns is a key unmet need in the pediatric population. Although actigraphy, a non-invasive, wrist-worn device, can be used to monitor sleep, sensor placement may cause skin irritation, which can further disrupt sleep in patients with AD [ 2 , 12 ]. There are currently no standardized scoring rules for actigraphy, a key limitation due to the absence of validation studies [ 13 ]. Furthermore, previous results in pediatric patients with AD show a positive correlation between self-reported sleep disturbance and itch but no correlation between itch and objective sleep measures such as sleep efficiency and sleep onset latency, as measured by actigraphy [ 14 ]. Other techniques have since been developed to generate objective measures of sleep, including the ADAM sensor, a flexible, wireless sensor placed on the dorsum of the hand, which can identify scratching based on acoustic signals [ 15 ]; piezoceramic bed-mounted devices, which detect scratching-related vibrations; and strain gauges, which measure finger flexion [ 16 ]. While some may show comparable accuracy to actigraphy, most of these devices have only been tested in healthy, non-pruritic adults and use technology that is costly or not widely available [ 16 ]. Improvements have been made in assessing sleep using advanced machine learning algorithms that monitor sleep and nocturnal breathing by measuring radio waves, such as wireless signals reflected from sleeping people, without the need for a sensor or other body-worn trackers [ 17 , 18 ]. The Emerald touchless sensor (Emerald Innovations, Cambridge, MA) is a wireless sensor with a machine learning platform that can objectively monitor scratching and sleep patterns in the patient's home environment without interfering with their activities or movements and without requiring them to wear sensors on their body. In the current study, the sensor, hung on the wall of the home, measures changes in wireless signals caused by body movements in the environment, and machine learning algorithms analyze the changes and capture the patient's scratching, sleep, and sleep stages. The sensor and its machine learning models have been evaluated and compared against gold standards, such as polysomnography [ 17 , 18 ] for sleep, where it showed an accuracy of about 80% in identifying sleep cycles, and video cameras with human labels for scratching, demonstrating accurate detection of nighttime scratching [ 19 , 20 ], which also correlated with a reduction in sleep quality [ 20 ]. The use of the sensor to measure sleep in neurological diseases (e.g., Parkinson's [ 21 , 22 ], dementia) [ 23 , 24 ], immune-related conditions (e.g., systemic lupus erythematosus) [ 25 ], chronic conditions (e.g., endometriosis) [ 26 ], and rare diseases (e.g., facioscapulohumeral dystrophy) [ 27 ] has been reported in the scientific literature. However, validation of the Emerald sensor specifically in pediatric populations with AD had not been established before this study. This paper reports a pilot study using the Emerald touchless sensor in pediatric patients with AD enrolled in the ongoing Pediatric Study in Atopic Dermatitis (PEDISTAD) registry [ 28 ]. The objective of this publication is to report the feasibility of using the Emerald touchless sensor to quantify nighttime scratching and sleep patterns in pediatric patients with AD (aged 6–11 years) enrolled in PEDISTAD who received standard-of-care (SoC) therapies.

Below is the link to the electronic supplementary material. Supplementary file1 (PDF 1054 KB) Supplementary file1 (PDF 1054 KB)