MedPath

A Deep Learning Method to Evaluate QT on Ribociclib

Recruiting
Conditions
Ribociclib
Breast Cancer
Interventions
Other: Acquisition of a digitized ECG by four modalities within 20 minutes
Registration Number
NCT05623397
Lead Sponsor
CMC Ambroise Paré
Brief Summary

"Deep-learning" is a fast-growing method of machine learning (artificial intelligence, AI) which is arousing the interest of the scientific committee in many medical fields. These methods make it possible to generate matches between raw inputs (such as the digital signal from the ECG) and the desired outputs (for example, the measurement of QTc). Unlike traditional machine learning methods, which require manual extraction of structured and predefined data from raw input, deep-learning methods learn these functionalities directly from raw data, without pre-defined guidelines. With the advent of big-data and the recent exponential increase in computing power, these methods can produce models with exceptional performance. The investigators recently used this type of method using multi-layered artificial neural networks, to create an application based on a model that directly transforms the raw digital data of ECGs (.xml) into a measure of QTc comparable to those respecting the highest standards concerning reproducibility.

The main purpose of this trial is to study the performance of our DL-AI model for QTc measurement (vs. best standards of QTc measurements, TCM) applied to the recommended ECG monitoring following ribociclib prescription for breast cancer patients in routine clinical care. The investigators will acquire ECG with diverse devices including simplified devices (one/three lead acquisition, low frequency sampling rate: 125-500 Htz) to determine if they'll be equally performant versus 12-lead acquisition machine to evaluate QTc in this setting.

Detailed Description

Not available

Recruitment & Eligibility

Status
RECRUITING
Sex
Female
Target Recruitment
70
Inclusion Criteria
  • Adult female patients requiring start of ribociclib based therapy for a breast cancer in their standard of care, as per their summary of product characteristic's indications
  • Association with hormone-based therapy in combination is authorized (aromatase inhibitors or fulvestrant)
  • Able to provide an informed consent
Exclusion Criteria
  • Any allergy or contra-indication to ribociclib as mentioned in their as summary of product characteristic's
  • Patients presenting a condition precluding accurate QTc measurements on electrocardiogram, i.e paced ventricular rhythm, multiples premature ventricular or supra-ventricular contractions, ventricular tachycardia, supraventricular arrhythmia (including atrial fibrillation, flutter or junctional rhythm)
  • Patients with an atrial pacing and sinus dysfunction
  • Patients presenting a contra-indication for ECG measurement, or with a device rendering ECG measurements impossible (i.e. Diaphragmatic pacing)
  • Patients presenting a contra-indication to ribociclib start; including association with prohibited drug potentializing the risk of TdP

Study & Design

Study Type
OBSERVATIONAL
Study Design
Not specified
Arm && Interventions
GroupInterventionDescription
Breast cancer patients administered ribociclib.Acquisition of a digitized ECG by four modalities within 20 minutesProspective cohort of consecutive breast cancer patients requiring ribociclib for their standard of care at the clinically indicated dose, as per treating physician prescription (600mg to 200mg/day for 21 days per 28 days cycle). Association with other hormone-derived therapeutics will be allowed.
Primary Outcome Measures
NameTimeMethod
Compare the values of QTc generated by method 1 (overlap method on triplicate of 10 seconds ECG concatenated, TCM; the method of reference) versus method 2 relying on AI methodology in patients' candidate for ribociclib startOne visit the day of ribociclib start (before ribociclib intake)

Comparison of the 2 methods (TCM vs. DL-AI) to demonstrate if there is a clinically relevant mean QTc difference ≥ 5msec between the 2 methods.

Secondary Outcome Measures
NameTimeMethod
Compare the values of QTc generated using method 2 (DL-AI) in patients' on/off ribociclib using a miniaturized and/or simplified ECG acquisition device (QT-Medical®, AliveCor®, a holter system (CGM HI-patch) versus using a digitized 12-lead acquisitionOne visit at baseline before ribociclib start and then day 14+/-3 and day 28+/-3 after ribociclib start

Compare QTc values obtained by DL-AI on/off ribociclib using a standard digitized 12-lead acquisition device (WELCH-ALYN ELI-280) versus each of three other miniaturized and/or simplified ECG acquisition devices (QT- Medical®, AliveCor®, CGM HI-patch®).

Compare the values of QTc generated by method 1 (overlap method after triplicate concatenation, TCM) versus method 2 (DL-AI) in patients' on/off ribociclib using a digitized 12-lead acquisition ECG deviceOne visit at day 14+/-3 and day 28+/-3 after start of ribociclib

Bland-Altman plots and intra-class correlation will be generated to compare QTc values obtained by TCM vs. DL-AI on ribociclib (Day 14+/-3 days after start) and off-ribociclib (Day 28 +/-3 of ribociclib cycles).

The clinico-demographic predictors of amplitude of QTc prolongation on ribociclib.One visit at baseline before ribociclib start and then day 14+/-3 and day 28+/-3 after ribociclib start

Nonlinear mixed models will be used to study clinico-demographic determinants associated with magnitude of QTc prolongation on ribociclib.

Learn ECG features at baseline using deep-learning predictors of magnitude of QTc prolongation on ribociclibOne visit at baseline before ribociclib start and then day 14+/-3 and day 28+/-3 after ribociclib start

Using deep-learning seeking for a model using ECG raw data at baseline to predict magnitude of QTc prolongation on ribociclib

Trial Locations

Locations (4)

Groupe Ambroise Paré, Hartmann

🇫🇷

Neuilly-sur-Seine, France

CIC - Hôpitaux Universitaires Pitié Salpêtrière, Paris, FRANCE

🇫🇷

Paris, France

Institut Gustave Roussy

🇫🇷

Villejuif, France

Hôpital Tenon

🇫🇷

Paris, France

© Copyright 2025. All Rights Reserved by MedPath