OPTimising a Screening Program to Detect Pacemaker-associated Heart Failure Using Artificial Intelligence

Overview

Pacemakers are an effective treatment for slow heart rates which improve symptoms and save lives. However, for some people pacemakers can cause heart failure (HF) because of the unnatural way in which they stimulate heart beats. In several studies conducted in West Yorkshire we showed that \~1/3 of patients with pacemakers have undiagnosed HF. We also showed that where HF is discovered, treating it with safe and inexpensive medications reduces the chances of being admitted to hospital or dying. However, detecting HF requires an echocardiogram (a heart ultrasound scan) which takes \~45 minutes, requires a skilled technician, and costs £120; or, to put it another way \~£540,000 to assess the \~4,500 patients cared for at our hospital. A new approach is needed. We think that using new technologies can improve our ability to screen for HF in people with pacemakers. We will test two approaches. First, we will assess whether a hand-held echocardiogram can measure heart function using artificial intelligence (AI) as accurately as a standard echocardiogram done by a skilled technician. Second, we will assess whether a finger-prick blood test can detect the presence of abnormal function as accurately as a standard echocardiogram.

Pacemakers are an effective treatment for slow heart rates which improve symptoms and save lives. However, for some people pacemakers can cause heart failure (HF) because of the unnatural way in which they stimulate heart beats. In several studies conducted in West Yorkshire we showed that \~1/3 of patients with pacemakers have undiagnosed HF. We also showed that where HF is discovered, treating it with safe and inexpensive medications reduces the chances of being admitted to hospital or dying. However, detecting HF requires an echocardiogram (a heart ultrasound scan) which takes \~45 minutes, requires a skilled technician, and costs £120; or, to put it another way \~£540,000 to assess the \~4,500 patients in Leeds. A new approach is needed. We think that using new technologies might improve our ability to screen for HF in people with pacemakers. Hand-held devices which are capable of cardiac imaging, and measurement of heart failure using Artificial Intelligence (AI) are now commercially available. However, these devices have not been specifically developed to detect heart dysfunction in people with pacemakers. Pacemakers result in a specific form of heart muscle weakness which is different to that caused by other conditions which result in HF, therefore it is essential to prove its accuracy before use in routine clinical practice. Another possible way to detect heart muscle weakness is by first screening patients with a blood test. N-terminal B- type natriuretic peptide (NT-proBNP) is a peptide released by the heart in response to strain or stress, and usually elevated in people with heart failure. Devices are now commercially available which can provide a measurement of NT-proBNP from a finger prick blood test, and crucially within a few minutes so that the result might be known during a patients routine appointment. Whilst this blood test cannot provide a diagnosis, it might help identify individuals who are at higher risk and should go on to have a standard echocardiogram. To make participation as simple as possible, and to avoid unnecessary visits to the hospital, patients will be approached prior to their routine pacemaker clinic appointment. Patients who have pacemakers attend the hospital every 6-12 months in person to undergo a check of their pacemaker's battery life and function. All potential participants will be identified by a member the patient's direct care team according to the study inclusion criteria. We will recruit patients with pacemakers and right ventricular pacing burden (the percentage of heat beats which are due to the pacemaker) ≥20% - this is important as we know that the risk of heart muscle weakness is correlated with the percentage of paced beats, with ≥20% being an accepted cut-off supported by our own data. We will exclude patients unable or unwilling to participate. We will also exclude patients with cardiac devices which mitigate against the risk of heart muscle weakness, or those likely to have been implanted for the management of known heart failure (conduction system pacemakers, cardiac resynchronisation therapy pacemakers/defibrillators, implantable cardioverter defibrillators, leadless pacemakers, single chamber atrial pacemakers). Potential participants will be sent an Invitation Letter and the Patient Information Sheet, offering them the opportunity to attend an appointment in the Leeds Cardiovascular Clinical Research Facility instead of their usual pacemaker clinic appointment. This will then be followed up by a telephone call from the direct care team, and only those who are potentially interested and confirm their agreement to have their information passed to the research team will be approached to arrange an appointment. On the day of the appointment, a member of the research team will approach the patient, explain the study and offer the opportunity to ask questions. Participants will have already had at least one week to read the Participants Information Sheet and consider whether they wish to take part. If willing to participate, they will then provide written consent to undertake the study procedures. All patients, regardless of whether they consent to participate, or not, will have a pacemaker safety check by an accredited cardiac physiologist. If patients agree to participate their appointment would last around one hour, for comparison a routine pacemaker appointment lasts around 20 minutes. Following the appointment their next appointment with their usual care team will be scheduled. Generic study procedures: Participants will undergo a routine device interrogation by an accredited cardiac physiologist as per usual clinical care to confirm device functionality and right ventricular pacing burden ≥20%. For participants who consent to participate the research team will collect demographic and clinical information from their electronic healthcare record. Participants will then undergo point-of-care blood testing for NTpro-BNP, a hand-held echocardiogram and a standard cart-based echocardiogram. Point-of-care NTpro-BNP blood testing: LumiraDx (LumiraDx Ltd, Cayman Islands) is a quantitative point-of-care, fingerpick blood test capable of measuring various biomarkers using a micro-sample of blood obtained from a finger prick. NT-proBNP is a routine blood test used to diagnose and monitor heart failure. The LumiraDx can measure NTpro-BNP with only 20uL of blood, is highly portable, with a turn-around time of only 12 minutes - feasible within the time constraints of a standard pacemaker follow-up appointment. For the patient, this entails having a small quantify of blood taken from a finger prick. The process results in very minimal discomfort and is almost identical to blood sugar testing patients with diabetes are requiring to do several times per day. Afterwards a small amount of cotton wool is placed on the finger tip and pressure applied. Their technology has been validated in a study of 659 samples compared to gold-standard commercially available laboratory equipment (Roche Cobas Elecsys proBNP II assay) and found to have comparable accuracy (slope 0.86, intercept -4.12, r 0.97). Hand-held cardiac ultrasound: participants will undergo a focussed hand-held echocardiogram with AI derived LVEF using the Kosmos platform (EchoNous Inc., Redmond, USA) a point-of-care, hand-held ultrasound device capable of cardiac imaging, which uses US2.AI (Ultrasound Anyone Anywhere, Singapore) integrated AI driven software capable of automated measurement of LVEF. The Kosmos platform assists the operator to obtain optimal images, prompts the image to be acquired when the image has been optimised, and provides an immediate measurement of LVEF. Existing data suggest the Kosmos platform is able to reliably detect abnormal LVEF in patients referred with suspected HF, however there are no data describing its use in a population with pacing-induced left ventricular dysfunction. Hand-held echocardiography will be done by a specialist nurse or cardiac physiologist who does not hold accreditation in echocardiography. They will have previously received focussed training in obtaining the images required, which will be deliberately limited (≤1 day) so as to ensure our findings are generalisable to teams without experience of echocardiography. Standard cart-based cardiac ultrasound: participants will undergo a transthoracic echocardiogram by a senior cardiac physiologist accredited by the British Society of Echocardiography (or equivalent international body) using a GE Healthcare Vivid E95 Ultra Edition Cardiac Ultrasound. For the participant this involves lying semi-reclined on a couch, with ultrasound images obtained from four locations on the chest wall. The scan takes around 25 minutes to obtain the images (with an additional 20 minutes to take measurements and prepare the report) and poses no risk to the patient. A minimum dataset will be obtained with LVEF measured by the biplane method where possible, as well as assess for any (at least) moderate valvular disease or other finding which could result in a change to patient management. This is a cross-sectional study, participants will be assessed at a single point in time. Provided there are no abnormal findings, participants will then be booked to attend their next scheduled pacemaker follow-up appointment as planned. Any serious adverse event (SAE) occurring in a research participant, where, in the opinion of the Chief Investigator the event was related to the administration of any of the research procedures will be reported to the Sponsor (governance- ethics@leeds.ac.uk) within one working day of the research team's awareness. SAE will be recorded during the participant's visit, and documented in the electronic healthcare record. Untoward events which occur after the study visit has been completed or are not directly related to the study visit will not be reported. It is anticipated that during this study we will identify important heart muscle weakness, or other abnormal findings on cardiac ultrasound. These findings, and any untoward event which occur as a consequence of new diagnoses/findings (such as hospitalisation) will not be reported to the Sponsor of Research Ethics Committee as these are regarded as being part of the patient's usual clinical care.