Novel Sensors and Artificial Intelligence for Detection of Acute Pulmonary Exacerbations in Cystic Fibrosis and Bronchiectasis

Primary Research Question Can novel sensors and artificial intelligence (AI) be used to detect acute pulmonary exacerbations (APEs) before becoming clinically apparent in patients with cystic fibrosis (CF) or non-cystic fibrosis bronchiectasis (non-CF bronchiectasis)?

Background Bronchiectasis is a clinical syndrome defined by abnormal, usually permanent dilatation of the bronchi associated with cough and sputum production. In addition, for many patients the daily symptoms of the disease are accompanied by unpredictable worsening of symptoms greater than the normal day-to-day variation and referred to as Acute Pulmonary Exacerbations (APEs).

CF is a life limiting, inherited condition that causes a build-up of thick, sticky secretions in the lungs (and other organs) that leads to severe and persistent bacterial infections. CF is characterised by periods of relative stability punctuated by abrupt clinical deterioration, known as APEs, which drive pulmonary inflammation, progressive lung damage and premature death.

Symptoms of APEs include change in sputum volume or colour, increased cough, increased dyspnoea, malaise, fatigue, or lethargy, temperature above 38°C, anorexia or weight loss. Change in physical examination of the chest, decrease in pulmonary function and / or radiographic changes indicative of pulmonary infection may also be seen.

APEs have a significant impact on patient outcomes. Severe APEs have been linked to accelerated lung function decline and patients with frequent APEs, particularly those experiencing three or more APEs per year, have a worse quality of life, more frequent hospitalisations, and increased mortality over 5 years. APEs are also a major contributor to healthcare cost.

Patients with CF and non-CF bronchiectasis can have a large treatment burden including medications and chest physiotherapy. This can take hours out of the day impacting on home and work life. In summary, the main goal of treatment is to reduce APE frequency and to detect exacerbations earlier.

Aims The aim of this study is to explore the potential use of a range of novel sensors to assist or improve the detection of health conditions in people with CF or non-CF bronchiectasis. The emphasis is on identifying onset and progression of APEs.

Rationale Current diagnosis of an APE relies heavily on patient-reported changes in baseline symptoms and accompanying declines in spirometry parameters. People with CF (PwCF) and those with non-CF bronchiectasis, however, invariably experience day-to-day symptom variation. For both the patient and the clinician, distinguishing these normal perturbations from that heralding an APE is vital to the preservation of lung health.

Using home monitoring data from the team's previous national feasibility study SmartCareCF (REC 14/EE/1244) in which PwCF monitored health at home (e.g., spirometry, pulse oximetry, activity, heart rate, etc.) in combination with self-reported survey responses (e.g., general wellness, cough, sleep quality) a prototype Artificial Intelligence (AI)-powered predictive algorithm that can forecast the likelihood of an impending acute pulmonary exacerbation (APE) an average of 10 days earlier than conventional care was created.

Previous studies (Telemed; REC12/EE/0462, Project Breathe; REC 18/NI/0215 and SmartCare-CF) have highlighted:

1. The importance of high frequency monitoring in order to implement AI based predictive algorithms that identify imminent clinical deteriorations

2. The need to move from 'active' to 'passive' sensing technology to improve ease of data collection and reduce treatment burden for patients.

   The aim now is to reduce patient burden further and investigate new technologies for detection of APEs by trialling a range of novel sensors to ascertain if a change in signal from start to end of an APE is shown. Then only one or more of these sensors to detect APE onset and to monitor response to treatment rather than numerous pieces of home monitoring equipment that is currently may be required. A number of novel sensors, named OMED (Owlstone), ARIA and ANNE (Sibel), that measure physiological parameters for example, heart rate, temperature, cough will be trialled.

   Study Design This is a single centre, observational, prospective cohort, feasibility, umbrella study. It is designed to run as non-disruptive with no impact on routine clinical practice. Patients with either CF or non-CF bronchiectasis will be enrolled from the Lung Defence and Cystic Fibrosis specialist centres at Royal Papworth Hospital NHS Foundation Trust (RPH).

   In total, up to 50 participants will be recruited over a 2-year period.

   The novel sensor will be trialled to see if there is a change in signal from the start of APE, and intravenous (IV) antibiotic treatment, and back to baseline. Participants may trial one or more novel sensor. Participants will be given the choice of which novel sensors to trial (if eligible).

   Participants will wear or use the novel sensor daily during the course of IV antibiotic treatment for APE. Participants will also be asked to record coughing and wellness scores daily on a 11-point Likert scale, either electronically or on paper depending on participant preference. Coughing is scored from 0-10 (0 no coughing, 10 maximal coughing) and wellness scored 0-10 (0 least well, 10 most well).

   If another APE should occur in the two weeks following IV antibiotic conclusion and a further course of IV antibiotics are required, then the sensor will continue to be used or worn until the conclusion of IV antibiotics.

   Primary Outcome The change in physiological parameter measures (sensor dependant) between the start and end of IV antibiotics for APE.

   Secondary Outcome The change in coughing and wellness scores between the start and end of IV antibiotics for APE.

   Primary Endpoint Mean absolute change in physiological parameter (sensor dependant) from start to end of IV antibiotics.

   Secondary Endpoints

   • Change in lung function as forced expiratory volume in one second predicted (FEV1%) between first and last day of IV antibiotics.
   • Change in coughing and wellness scores between first and last day of IV antibiotics. Coughing and wellness scores are subjectively scored by participants on a 11-point Likert scale. For coughing, the higher the score the more the participant is coughing. For wellness, the higher the score the more well the participant feels.

   Sample Size and Data Analysis Sample size will vary depending on the sensor being tested. These novel sensors have not been tested in the CF or non-CF bronchiectasis population before. The aim is to recruit sufficient participants to ensure the utility of the sensor has been tested balanced with the high cost of it can be determined if larger studies may be warranted.

   Statistical Analysis All data collection i.e. physiological parameters such as heart rate will be jointly analysed using a variety of statistical and machine learning techniques. Participants will be allocated an anonymised ID for the study. No identifiable information will be shared outside the core NHS clinical research team.

   The data generated by the study will be analysed as follows:

   • Link-anonymised physiological data will be analysed by University of Cambridge and the sensor company (with Data Sharing Agreement in place).
   • Link-anonymised clinical data will be analysed by the clinical research team, RPH and University of Cambridge and the sensor company (with Data Sharing Agreement in place).
   • Link-anonymised coughing and wellness data will be analysed by the clinical research team RPH and University of Cambridge the sensor company (with Data Sharing Agreement in place).

   The methodologies for statistical analysis will include standard multivariate analysis, linear and non-linear regression analysis.

   Recruitment Recruitment will occur over a 2-year period and participants will remain in the study for up to 5 weeks (unless repeat or prolonged courses of IV antibiotics are required). APEs will be defined using the Hill Criteria

   Hill Criteria

   A person with bronchiectasis with a deterioration in three or more of the following key symptoms for at least 48 hours:

   1. Cough
   2. Sputum volume and/or consistency
   3. Sputum purulence
   4. Breathlessness and/or exercise tolerance
   5. Fatigue and/or malaise
   6. Haemoptysis AND a clinician determines that antibiotic treatment is required

   Consent Process and Visit Schedule Study Visits Each participant will be assigned a study identifier and number. The study numbers will be allocated sequentially. Each participant will have an anonymous email account to enable anonymity when using the sensor app.

   The clinical research team will record the following participant specific study information on REDCap, from routine clinical care records during enrolment.

   • Date commenced on the study
   • Consent date and status
   • Date of birth
   • Age
   • Biological sex at birth
   • Ethnicity
   • Height at the start of the study
   • Weight at the start of the study
   • Episodes of APE will be recorded and symptoms as per the Hill Criteria will also be documented.
   • FEV1 % predicted and L at the start and end of APE IV antibiotic treatment.
   • Microbial status at the start of the study
   • Regular medications at the start of the study
   • Microbiological cultures results during study
   • Number of IV antibiotic days during study.
   • Number of days of oral antibiotics (from clinical care records and participant reported) during the study.
   • Usual APE treatment at the start of the study
   • C-reactive protein (CRP) and white blood count (WBC) levels from start and end of APE IV antibiotic treatment.

   Height, weight, FEV1 % predicted and L, CRP and WBC may be noted from those taken from up to 2 weeks prior to the participant commencing on the study if not available on the start date.

   FEV1 % predicted and litres (L), C-reactive protein (CRP) and white blood cell count (WBC) may be taken from those recorded up to 2 weeks post the participant ending the study, if not available on the end date.

   If there were no spirometry results obtained within the 2-week pre- and / or post- study period, the clinical research team can manually conduct this test to collect spirometry results.

   • Blood glucose levels (from random blood glucose, oral glucose tolerance test or anonymised continuous glucose monitoring report) from any recorded during the 2 weeks prior to APE treatment and those recorded during APE treatment.
   • The number of contacts for technical issues with equipment, during the study The above parameters are routinely collected as part of normal clinical care.

   All participants will be shown how to set up and use the sensor and how to record participant coughing and wellness scores.

   Informed Consent Procedure

   • Written informed consent will be obtained by a member of the clinical team or clinical research team after a suitable time has elapsed during which the patient has had ample time (minimum 1 hour) to read the patient information sheet, consider the trial and ask any questions.
   • Each subject will be informed that participation in the study is voluntary and that withdrawal of consent will not affect subsequent medical treatment.
   • Subjects may be replaced if withdraw (or are withdrawn) during the first 4 weeks of participation on the study. Participants may be withdrawn due to medical reasons by the Investigator.
   • Participants will be contacted by telephone / email or during hospital attendance (face to face or virtual) depending on participant preference by the clinical research team during the study. This will be to check that recording coughing and wellness scores are being recorded daily and no issues with using the sensor daily.

   Follow up Visits Participants do not need to attend study specific follow ups.

   Data Collection Form Completion The Investigator should ensure the accuracy, completeness, legibility and timeliness of the data recorded in the data collection forms (DCFs) and in all required reports to e.g. the Sponsor, Funder, Research and Development (R \& D) and Research Ethics Committee.

   Source Documentation

   • The clinical study data is being collected and managed using REDCap (Research Electronic Data Capture) electronic data capture tools hosted at University of Cambridge. REDCap is hosted on the Clinical School Secure Data Hosting Service (SDHS) which provides an ISO:27001 certified 'Safe Haven'. Only N3 compliant network providers will be used, and access will be password protected.
   • Every participant in the study will be given a Unique Identifier (UID). All study assessment data will be stored separately from personally identifiable data and referenced via the UID to maintain anonymity of the data accessible for research purposes.

   Monitoring and Audit

   • The project data will be monitored by the Research Officer in accordance with Trust policy. For this study the first ten consent forms will be monitored and further will be monitored only if it is deemed necessary. The project data will be monitored by the Research Officer quarterly.

   Adverse and Serious Adverse Events As these studies are observational and non-interventional and do not meet the HRA criteria that would require serious event reporting none will be reported in this study, except for adhesive monitoring for Sibel.