The use of artificial intelligence and mobile health technologies to identify patients at the highest risk of atrial fibrillation (irregular and often abnormally fast heart rate)
| ISRCTN | ISRCTN17993837 |
|---|---|
| DOI | https://doi.org/10.1186/ISRCTN17993837 |
| IRAS number | 293493 |
| Secondary identifying numbers | IRAS 293493, CPMS 50698 |
- Submission date
- 02/11/2021
- Registration date
- 22/11/2021
- Last edited
- 13/04/2022
- Recruitment status
- Suspended
- Overall study status
- Suspended
- Condition category
- Circulatory System
Plain English summary of protocol
Background and study aims
Atrial fibrillation (AF) is the most common heart rhythm disturbance (an irregular and often abnormally fast heart rate). Individuals with atrial fibrillation have an increased risk of developing strokes. AF can be persistent or paroxysmal (intermittent). People with AF can be unaware that they have the condition as they may not display any symptoms. If AF is identified earlier, strokes can be prevented with treatment in the form of a blood thinner (known as anticoagulation). In order to make a diagnosis, a heart tracing called an electrocardiogram (ECG) will be required.
Currently, over 100 individuals will need to be screened in order to pick one individual with AF. However, with our AF risk prediction machine learning algorithm, we predict that we can increase our yield in identifying patients at risk of developing atrial fibrillation. The AF risk prediction machine learning algorithm was developed from previous AF risk score models and primary care data. Early data has suggested that it reduces the number needed to screen to nine to pick one patient with atrial fibrillation.
There has been a recent surge in the number of technologies available for heart rhythm monitoring. As more people have smartphones, this provides us with more opportunities to monitor patients for longer and more frequently and ultimately, diagnose AF earlier. We would like to compare these technologies for AF detection among individuals identified as being at risk of atrial fibrillation by the AF risk prediction machine learning algorithm.
Aims:
1. To evaluate the potential benefit of using our machine learning algorithm to aid AF screening.
2. To compare the use of different health technologies for AF screening
3. To evaluate the optimal duration and frequency of rhythm monitoring for AF screening
Who can participate?
Participants who have been identified to be at risk of atrial fibrillation by the AF risk prediction machine learning algorithm. This algorithm will be run on general practice databases in the Hounslow primary care networks to generate a list of potential participants. They must not have a diagnosis of AF already and they must not have a cardiac electronic implantable device to be eligible to participate in the study.
What does the study involve?
The study involves screening the above participants for AF. All participants will have a single lead ECG using an AliveCor device with a Kardia App. If no AF is identified, the participants will undergo further monitoring for a period of 3 months. The participants will be divided into four groups by cluster randomisation of the general practices. The groups are as follows:
1. ECG group
This group will not have further monitoring after their single lead ECG.
2. Photoplethysmography group
This group will receive a Fibricheck App for their smartphones. They will need to take photoplethysmography recordings with their smartphone twice daily for 3 months. The recordings will be reviewed for the detection of atrial fibrillation.
3. AliveCor group
This group will receive an AliveCor device and a Kardia App for their smartphones. They will need to take single lead ECG recordings with their smartphone twice daily for 3 months. The recordings will be reviewed for the detection of atrial fibrillation.
4. Holter monitoring group
This group will receive 3 Holter monitors/ ECG patches over 3 months. The first monitor will be worn for 72 h. The subsequent monitors will be 24 h. There will be a 6-week interval between monitors.
If AF is identified at any point, the participant will be referred to their clinical team for appropriate treatment.
What are the possible benefits and risks of participating?
The possible benefits of participating are earlier identification of atrial fibrillation and earlier treatment for atrial fibrillation and therefore possible prevention of further AF related strokes or conditions. However, it is also possible there will be no direct benefit to the research participant. There are no foreseeable risks of participating in this study.
Where is the study run from?
West Middlesex University Hospital (UK)
Hounslow Clinical Commissioning Groups
When is the study starting and how long is it expected to run for?
May 2020 to June 2023
Who is funding the study?
Bristol Myers Squibb Pharmaceutical Limited (UK)
Chelsea and Westminster plus charity (UK)
Who is the main contact?
Dr Pavidra Sivanandarajah (Pavidra.sivanandarajah1@nhs.net)
Contact information
Scientific
Cardiology Department
West Middlesex University Hospital
Chelsea and Westminster NHS Foundation Trust
Twickenham Road
Isleworth
London
TW7 6AF
United Kingdom
 ORCID ID |
0000-0002-6108-4450 |
|---|---|
| Phone | +44 (0)2083215336 |
| pavidra.sivanandarajah1@nhs.net |
Study information
| Study design | Prospective observational cohort study |
|---|---|
| Primary study design | Observational |
| Secondary study design | Cohort study |
| Study setting(s) | GP practice |
| Study type | Screening |
| Participant information sheet | 40618 AMLA-AF PIS alivecor v1.1 25Oct2021.pdf |
| Scientific title | Application of machine learning algorithm to identify patients at highest risk of atrial fibrillation for targeted screening |
| Study acronym | AMLA-AF |
| Study objectives | Current hypothesis as of 29/11/2021: The AF risk prediction machine learning algorithm can be used to identify patients at the highest risk of atrial fibrillation (AF) for targeted screening. _____ Previous hypothesis: The PULsE AI machine learning algorithm can be used to identify patients at the highest risk of atrial fibrillation (AF) for targeted screening |
| Ethics approval(s) | Approved 01/12/2021, London- Bloomsbury Research Ethics Committee (HRA RES Centre Manchester, 3rd Floor Barlow House, 4 Minshull Street, Manchester M1 3DZ, UK; +44 0207 104 8196; bloomsbury.rec@hra.nhs.uk), ref: 21/LO/0709 |
| Health condition(s) or problem(s) studied | Atrial fibrillation (AF) |
| Intervention | Current intervention as of 29/11/2021: Participants will be identified using the AF risk prediction machine learning algorithm from the Hounslow primary care networks. The study involves screening the above participants for AF. All participants will have a single lead ECG using an AliveCor device with a Kardia App. If no AF is identified, the participants will undergo further monitoring for a period of 3 months. The participants will be divided into four groups by cluster randomisation of the general practices. The groups are as follows: 1. ECG group This group will not have further monitoring after their single lead ECG. 2. Photoplethysmography group This group will receive a Fibricheck App for their smartphones. They will need to take photoplethysmography recordings with their smartphone twice daily for 3 months. The recordings will be reviewed for the detection of atrial fibrillation. 3. AliveCor group This group will receive an AliveCor device and a Kardia App for their smartphones. They will need to take single lead ECG recordings with their smartphone twice daily for 3 months. The recordings will be reviewed for the detection of atrial fibrillation. 4. Holter monitoring group This group will receive 3 Holter monitors/ ECG patches over 3 months. The first monitor will be worn for 72 h. The subsequent monitors will be 24 h. There will be a 6-week interval between monitors. If AF is identified at any point, the participant will be referred to their clinical team for appropriate treatment. _____ Previous intervention: Participants will be identified using the PULsE AI machine learning algorithm from the Hounslow primary care networks. The study involves screening the above participants for AF. All participants will have a single lead ECG using an AliveCor device with a Kardia App. If no AF is identified, the participants will undergo further monitoring for a period of 3 months. The participants will be divided into four groups by cluster randomisation of the general practices. The groups are as follows: 1) ECG group This group will not have further monitoring after their single lead ECG. 2) Photoplethysmography group This group will receive a Fibricheck App for their smartphones. They will need to take photoplethysmography recordings with their smartphone twice daily for 3 months. The recordings will be reviewed for the detection of atrial fibrillation. 3) AliveCor group This group will receive an AliveCor device and a Kardia App for their smartphones. They will need to take single lead ECG recordings with their smartphone twice daily for 3 months. The recordings will be reviewed for the detection of atrial fibrillation. 4) Holter monitoring group This group will receive 3 Holter monitors/ ECG patches over 3 months. The first monitor will be worn for 72 hrs. The subsequent monitors will be 24hours. There will be a 6-week interval between monitors. If AF is identified at any point, the participant will be referred to their clinical team for appropriate treatment. |
| Intervention type | Device |
| Pharmaceutical study type(s) | |
| Phase | Not Applicable |
| Drug / device / biological / vaccine name(s) | Not applicable |
| Primary outcome measure | Detection of atrial fibrillation with a one-off ECG |
| Secondary outcome measures | There are no secondary outcome measures |
| Overall study start date | 01/05/2020 |
| Completion date | 30/06/2023 |
Eligibility
| Participant type(s) | All |
|---|---|
| Age group | Adult |
| Lower age limit | 18 Years |
| Sex | Both |
| Target number of participants | 1,800 |
| Key inclusion criteria | Current inclusion criteria as of 29/11/2021: 1. Aged 18 years or above 2. Identified as high-risk for AF by our 'AF risk prediction machine learning algorithm 3. Access to smartphone depending on allocated screening group _____ Previous inclusion criteria: 1. Aged 18 years old or above 2. Identified as high-risk for AF by our PULsE AI machine learning algorithm 3. Access to smartphone depending on allocated screening group |
| Key exclusion criteria | 1. Have already a diagnosis of atrial fibrillation prior to study enrolment 2. Below the age of 18 years old 3. Presence of cardiac electronic implantable device |
| Date of first enrolment | 11/04/2022 |
| Date of final enrolment | 30/11/2022 |
Locations
Countries of recruitment
- England
- United Kingdom
Study participating centre
Twickenham Road
Isleworth
London
TW7 6AF
United Kingdom
Sponsor information
Hospital/treatment centre
Research and Development Office
Unit G3
Harbour Yard
Chelsea Harbour
London
SW10 0XD
England
United Kingdom
| Phone | +44 (0)20 3315 6825 |
|---|---|
| damon.foster2@nhs.net | |
| Website | https://www.chelwest.nhs.uk/about-us/research-innovation-and-quality-improvement-riqi/research-development |
"ROR" |
https://ror.org/02gd18467 |
Funders
Funder type
Industry
Government organisation / For-profit companies (industry)
- Alternative name(s)
- Bristol-Myers Squibb Company, BMS
- Location
- United States of America
No information available
Results and Publications
| Intention to publish date | 30/06/2024 |
|---|---|
| Individual participant data (IPD) Intention to share | Yes |
| IPD sharing plan summary | Published as a supplement to the results publication |
| Publication and dissemination plan | Planned publication in a high-impact peer-reviewed journal |
| IPD sharing plan | All data generated or analysed during this study will be included in the subsequent results publication |
Study outputs
| Output type | Details | Date created | Date added | Peer reviewed? | Patient-facing? |
|---|---|---|---|---|---|
| Participant information sheet | Alivecor group version 1.1 |
25/10/2021 | 16/11/2021 | No | Yes |
| Participant information sheet | ECG group version 1.0 |
26/10/2021 | 16/11/2021 | No | Yes |
| Participant information sheet | Fibricheck group version 1.2 |
25/10/2021 | 16/11/2021 | No | Yes |
| Participant information sheet | Holter group version 1.1 |
25/10/2021 | 16/11/2021 | No | Yes |
| Protocol file | version 1.1 | 25/10/2021 | 16/11/2021 | No | No |
| HRA research summary | 28/06/2023 | No | No |
Additional files
- 40618 AMLA-AF Protocol v1.1 25Oct2021.pdf
- 40618 AMLA-AF PIS alivecor v1.1 25Oct2021.pdf
- Alivecor group
- 40618 AMLA-AF PIS ECG v1.0 26Oct2021.pdf
- ECG group
- 40618 AMLA-AF PIS fibricheck v1.2 25Oct2021.pdf
- Fibricheck group
- 40618 AMLA-AF PIS holter v1.1 25Oct2021.pdf
- Holter group
Editorial Notes
13/04/2022: The study has been suspended because the intervention requires modification.
07/02/2022: The following changes have been made:
1. The recruitment start date has been changed from 31/01/2022 to 11/04/2022.
2. The ethics approval date has been added.
10/01/2022: The intention to publish date has been changed from 30/06/2021 to 30/06/2024.
06/01/2022: The recruitment start date was changed from 10/01/2022 to 31/01/2022.
07/12/2021: Internal review.
03/12/2021: The recruitment start date has been changed from 01/12/2021 to 10/01/2022.
29/11/2021: The study hypothesis, intervention, participant inclusion criteria and plain English summary have been updated in order to replace "PULsE AI machine learning algorithm" to "AF risk prediction machine learning algorithm" throughout the record.
16/11/2021: Trial's existence confirmed by Bloomsbury REC.
