AUTHORS
Patrick Harrington, MD, Kim Summers2, Jay Corr3, Muna Yusuf4, Alfonso Portabales5, Debbie Street6, Mark Taylor7, Amna Sheikh8, Rebecca Frewin9, Alesia Khan, MD10, Samah Alimam, MD11, Claire Woodley12, Catherine Loughran, MRPharmS, BPharm13, Priya Sriskandarajah, MBBS, BSc, MRCP14, Ali Mahdi, MD15, Andrea Duminuco16, Shahram Kordasti, MD, PhD17, Orlando Agrippa18, Claire Harrison, DM19
AFFILIATIONS
1 Guy's and St Thomas' NHS Foundation Trust, London, United Kingdom
2 Sanius Health, London, United Kingdom
3 Sanius Health, London, United Kingdom
4 Sanius Health, London, United Kingdom
5 Sanius Health, London, United Kingdom
6 MPN Voice, London, United Kingdom
7 MPN Voice, London, United Kingdom
8 Guy's and St Thomas' NHS Foundation Trust, London, United Kingdom
9 Gloucester Hospitals NHS Foundations Trust, Gloucester, United Kingdom
10 Leeds Teaching Hospitals NHS Trust, Leeds, United Kingdom
11 University College London Hospital, London, United Kingdom
12 Guy's and St Thomas' NHS Foundation Trust, London, United Kingdom
13 University Hospitals Leicester, Nottingham, United Kingdom
14 Guy's and St Thomas' NHS Foundation Trust, London, United Kingdom
15 Aneurin Bevan University Health Board, Wales, United Kingdom
16 Guy's and St Thomas' NHS Foundation Trust, London, United Kingdom
17 Guy's and St Thomas' NHS Foundation Trust, London, United Kingdom
18 Sanius Health, London, United Kingdom
19 Guy's and St Thomas' NHS Foundation Trust, London, United Kingdom
ABSTRACT
Introduction:
Wearable technology and mobile applications provide the opportunity for continuous patient monitoring. Here we provide findings from the myMPNVoice app, launched in April 2024 in the U.K., focusing on biometric data from the associated wearable device.
Methods:
Participants with an MPN diagnosis enrolled by self-referral through an online platform. The app allows patients to record MPN-10 total symptom score (TSS) amongst other patient reported outcomes (PRO). Biometric data were collected from a Withings watch using accelerometery, which feeds into the app.
Electronic health record (EHR) data were received from primary and tertiary care settings. An augmented Large Language Model (LLM, Llama) was used to identify a concept of interest, and of their negation, before extracting relevant values. Retrieval Augmented Generation improved the LLM’s medical domain knowledge. Paired mean values for haematologic laboratory results (haemoglobin [Hb], white cell count [WCC], platelets) and biometric and PRO data, over a weekly period, were assessed.
A Linear Mixed Model was used to account for repeated measures, with age and sex as covariates.
Results:
526 patients are enrolled in the study (app and watch n=329, app only n=197) with data currently available from 473 patients. Of these 256 (49%) had ET, 166 (32%) had PV, 91 (17%) had MF and 13 (2%) MPN not otherwise specified (NOS). Patients had a median follow up of 300 days. Median age was 58 years (range 25-84), and 65% were female. A total number of 3.4 million data points across 74 distinct variables were collected. Of 264 patients on study for >12 months, 207 (78%) remained active (mean 15 data entries per month). Similarly, of 90 patients on study for >24 months 83 (92%) remained active (mean 16 data entries per month).
Men had significantly increased activity across multiple parameters including; steps, distance, active time, active and total calories as well as intense and moderate activity. There were no significant differences in sleep metrics based on sex. Surprisingly there were few significant differences in activity data based on age, however older patients had increased wake-up count and wake-up duration.
Unexpectedly, patients with MF recorded lower MPN-10 TSS compared with other MPN subtypes. However, MF patients simultaneously had lower activity metrics than ET and PV, including active duration, step count, distance, elevation and exercise of soft intensity. MF patients also had significantly lower total sleep duration and increased wake-up count, when compared with other diagnostic groups.
Significant correlations were observed between multiple PRO variables and biometric parameters across all MPN subtypes, with higher MPN-10 TSS associated with reduced activity across multiple metrics. Of note, fatigue, the most frequent symptom in MPN, was associated with reduced QoL, with higher fatigue score significantly correlated with lower psychological wellbeing, worse mobility and self-care and increasing difficulty with usual activities.
Regarding the relevance of different biometric data, firstly, activity variables were more closely associated with symptoms than sleep metrics. Of the different activity variables, step count was the most relevant metric with greatest, inverse, correlation with total MPN-10 TSS and inactivity as well as strongest positive correlation with QoL and psychological wellbeing. Of the sleep parameters, total sleep duration broadly showed closest alignment with various symptom metrics.
Regarding laboratory results, values were extracted from EHR in 223 patients. There was greater alignment with biometric data than with patient reported symptom data. In ET, there was significant inverse correlation between platelet count and steps/distance, whereas no correlation with symptoms was observed. Similarly, for WCC in ET there was inverse correlation with active time, distance, steps and calorie expenditure but no association with symptoms. Finally, for Hb, there was again, inverse correlation with steps/distance in ET patients but no correlation with MPN-10 TSS.
Conclusion:
There continues to be concern around use of symptoms as a primary end point for MPN clinical trials. In this study we demonstrate that biometric activity data may be superior with regards to providing objective monitoring. We continue to show broad correlation between activity and symptoms, however, now also show that biometrics are more aligned with laboratory results.
DISCLOSURES
Alesia Khan, MD: No relevant conflicts of interest to declare.