AI Algorithms in Prediction of ACS Based on Leukocyte Properties

Acute Coronary Syndrome Clinical Trial

Official title:

Diagnostic Performance of Artificial Intelligence Algorithms in Prediction of Acute Coronary Syndrome Based on White Blood Cell Properties (AI-ACS Trial)

Clinical Trial Details — Status: Recruiting

Administrative data

• NCT number: NCT06384846
• Other study ID #: AI ACS
• Status: Recruiting
• Start date: February 1, 2024
• Est. completion date: December 31, 2026

Study information

• Verified date: April 2024
• Source: RobotDreams GmbH
• Contact: Dimitrij Shulkin, M.Sc.
• Phone: +43-676-5150578
• Email: shulkin[@]robotdreams.co
• Is FDA regulated: No
• Study type: Observational

Clinical Trial Summary

The goal of this observational study is to find out if artificial intelligence (AI) can accurately predict acute coronary syndrome (ACS) using data on white blood cells in adults. The main question it aims to answer is: - Can AI algorithms based on white blood cell data predict ACS with accuracy comparable to that of high-sensitivity cardiac troponin (hs-cTn)? Researchers will look at how the AI model's predictions stack up against the standard hs-cTn blood tests to see which is more accurate in diagnosing ACS. Participants in this study will have already had blood tests as part of their usual care. Their previously collected health information and blood test results will be used to help train and test the AI algorithms. Participants will not undergo any new procedures for the study itself.

Description:

The AI-ACS clinical trial is an observational, single-center study aimed at assessing the diagnostic performance of AI algorithms that utilize white blood cell (WBC) data to predict acute coronary syndrome (ACS) in patients presenting with acute chest pain. This trial leverages advanced artificial intelligence to analyze high-dimensional measurements of WBC properties to improve the prediction and differentiation of ACS from other non-cardiac causes of chest pain, comparing these predictions against traditional high-sensitivity cardiac troponin (hs-cTn) measurements. Technical Description of the Study Protocol The AI-ACS trial uses a prospective, observational case-control design conducted at the Medical University of Graz. It is structured into two main phases: training and testing of AI models. WBC data are collected through routine blood tests performed upon patient admission, using the Sysmex XN series hematology analyzers. This data is used to train AI algorithms at regular intervals, aiming to refine their diagnostic accuracy over the course of the study. Quality Assurance and Registry Procedures The AI-ACS trial incorporates several measures to ensure the quality and integrity of the data collected, as well as adherence to standard operating procedures: Data Validation and Quality Assurance Plan: Continuous on-site monitoring and periodic audits are conducted to ensure adherence to the clinical protocol and regulatory compliance. Data validation procedures are implemented to check the accuracy and consistency of the WBC data entered into the study's database. Automated data checks compare new entries against predefined rules for range and consistency with other data fields. Source Data Verification: Source data verification is carried out by comparing the electronic data captured in the study database with original medical records and laboratory reports to assess the accuracy and completeness of the data. Data Dictionary: The study utilizes a detailed data dictionary that includes descriptions of each variable collected, including the source of the variable (e.g., patient demographics, laboratory results), coding information, and normal ranges. This dictionary helps maintain consistency in data interpretation and analysis. Standard Operating Procedures (SOPs): SOPs for patient recruitment, data collection, data management, analysis, and reporting are well-documented and followed throughout the study. These procedures include detailed steps for handling adverse events and changes in study protocol. Sample Size Assessment: The study is designed with a sample size of 2100 participants for training (700 per cohort) and 600 for testing (300 per cohort), calculated to provide sufficient power to detect significant differences in diagnostic performance of the AI models versus hs-cTn. Plan for Missing Data: Procedures to address missing or inconsistent data include imputation techniques and sensitivity analyses to evaluate the impact of missing data on study results. Statistical Analysis Plan: The statistical analysis plan outlines the methods used to evaluate the diagnostic performance of the AI models, including Receiver Operating Characteristic (ROC) curve analysis, area under the curve (AUC) calculations, and sensitivity and specificity assessments. Comparisons of AI models with hs-cTn measurements will be conducted using logistic regression models adjusted for potential confounders. Study Aims and Hypotheses The primary aim of the AI-ACS trial is to train and validate AI models capable of accurately predicting ACS from WBC data, potentially surpassing the diagnostic performance of standard hs-cTn assays. The hypothesis is that AI-driven analysis of WBC properties can more accurately differentiate between ACS and non-ACS causes of chest pain, thus improving clinical decision-making and reducing unnecessary medical interventions. By advancing how we utilize routine blood tests with AI, the AI-ACS trial seeks to enhance the rapid identification of patients at risk of ACS, thereby potentially transforming the standard of care in emergency cardiovascular diagnosis.

Recruitment information / eligibility

• Status: Recruiting
• Enrollment: 2700
• Est. completion date: December 31, 2026
• Est. primary completion date: July 31, 2026
• Accepts healthy volunteers: Accepts Healthy Volunteers
• Gender: All
• Age group: 18 Years and older

Eligibility

Inclusion Criteria:

• Male or Female, aged 18 years or above
• Participant is willing and able to give informed consent for participation in the study
• Subjects presenting without chest pain or with stable angina pectoris but without indication for revascularization during coronary angiography; identical evaluation results by review board required
• Exclusion of elevated hs-cTn
• Criteria for timing of blood sampling for collection of WBC and hs-cTn data need to be fulfilled (see 5.14) o Subjects with no or stable angina pectoris must have provided WBC data and at least one hs-cTn value any time before start of coronary angiography.
• Between initial blood sampling to collect WBC data and coronary angiography, the subject must not develop suspicion of ACS.

Exclusion Criteria:

• Age < 18 years old
• Subject refuses informed consent
• Collection of WBC and hs-cTn data is not possible
• Criteria for timing of blood sampling for collection of WBC and hs-cTn data cannot be fulfilled
• Suspicion of ACS occurred in subjects with no or stable angina pectoris any time between initial blood sampling and start of coronary angiography

Related Conditions & MeSH terms

• Acute Coronary Syndrome
• Angina Pectoris
• Infarction
• Myocardial Infarction
• NSTEMI - Non-ST Segment Elevation MI
• ST Elevation Myocardial Infarction
• STEMI - ST Elevation Myocardial Infarction
• Syndrome

Locations

Country	Name	City	State
Austria	Landeskrankenhaus-Universitätsklinikum Graz	Graz	Styria / Steiermark

Sponsors (1)

Lead Sponsor	Collaborator
RobotDreams GmbH

Country where clinical trial is conducted

Austria, 

Outcome

Type	Measure	Description	Time frame	Safety issue
Primary	Training of AI models	Diagnostic performance of AI models in predicting ACS, evaluated by area under curve (AUC) under the receiver operating characteristic (ROC) curve	36 months
Primary	Testing of AI models	Diagnostic performance of AI models in predicting ACS, evaluated by AUC under ROC curve; ; Specificity and sensitivity of AI models to predict ACS in subjects with suspected ACS, calculated from AUC under ROC curve	36 months
Secondary	Training of AI models	Sensitivity of AI models to predict ACS; ; Specificity of AI models to predict ACS	36 months
Secondary	Testing of AI models:	Sensitivity of AI models to predict ACS; Specificity of AI models to predict ACS; Sensitivity of hs-cTn to predict ACS; Specificity of hs-cTn to predict ACS; Combined sensitivity of AI models and hs-cTn to predict ACS; Combined specificity of AI models and hs-cTn to predict ACS; AUC under ROC curve of hs-cTn predicting ACS; AUC under ROC curve of AI models and hs-cTn predicting ACS; Difference in predicting ACS between hs-cTn and AI models using AUC under ROC curve	36 months