Advancement of precision health for neuromuscular disorders using effective patient registry and machine learning

Abstract

Neuromuscular disorders (NMDs) encompass conditions that result in disability and premature death. The high genetic variability and rareness of NMDs pose significant obstacles to conducting clinical studies. Recently, the development in patient registries and machine learning have propelled global clinical studies in NMDs, leading to the emergence of precision health. Precision health refers to an individualized approach to optimize clinical care by focusing on disease prevention, prediction, personalization and patient participation. In this thesis, I will explore different advancements of precision health for NMDs. We designed patient/clinician-reported questionnaires to collect demographic, clinical and genetic characteristics, family history and information on disease-modifying treatments (DMTs). The patient registry recruited participants from three Hong Kong University-affiliated neurology clinics at Hong Kong Children’s Hospital, Queen Mary Hospital and the Duchess of Kent Children's Hospital at Sandy Bay. We also developed a self-registration online platform. Furthermore, we collaborated with the Department of Electrical and Electronic Engineering to develop a machine learning model for facial weakness prediction. Since June 2019, the registry has recruited 237 patients with over 20 diagnoses. Most common were facioscapulohumeral muscular dystrophy (FSHD) (27.4%, n=65) and spinal muscular atrophy (SMA) (15.6%, n=37). A quarter of patients (79.3%, n=179) received a molecular diagnosis and 24.9% of patients (n=59) were undergoing DMTs. Both patient registry and machine learning facilitated research on facial weakness prediction in FSHD patients, achieving an accuracy rate of 87.1% in our pilot study. Furthermore, the FSHD registry facilitated the study of genotypic-phenotypic profiles and its correlation. With the genetic diagnosis achieved in 61 patients, we found moderate correlations between the number of D4Z4 repeat units and age of disease onset (ρ=0.426, p<0.001) as well as clinical severity (r=-0.556, p<0.001). Additionally, I led the establishment of the pioneering personalized respiratory and motor telerehabilitation program for general pediatric NMD patients. Not only we noticed an improvement in forced vital capacity and health-related quality of life (HRQOL) in most patients, but also the maintenance of most motor function. The program received satisfaction (4.00/5.00) and recommendations (4.38/5.00) from all patients and their families. Next, our SMA registry facilitated the safety and efficacy assessment of DMTs. Our findings demonstrated positive outcomes in terms of safety, motor function, HRQOL and parent/patient-reported outcomes for nusinersen and risdiplam. During the COVID-19 pandemic, the registry facilitated studies on HRQOL and vaccination. We demonstrated a poorer HRQOL in NMD children compared to typically-developed children, as measured by the Pediatric Quality of Life Inventory 4.0 Generic Core survey (55.7 vs 81.6, p<0.001). Furthermore, parental guidance on electronic device usage was identified as a predictive factor of HRQOL. Importantly, all patients seroconverted against the wildtype SARS-CoV-2 after receiving two doses of BNT162b2 or CoronaVac, including Duchenne muscular dystrophy patients on corticosteroid treatment. These data are vital in preventing the aggravation of psychological problems and severe COVID-19 complications in NMDs. In conclusion, my study has made significant advancements in healthcare for patients with NMDs. It helps to optimize clinical care in terms of disease prevention, prediction, personalization and patient participation.