At-Home Breath Data Collection for Signatures of Type 2 Diabetes: A Pilot Clinical Study

Area of Science:

• Biomedical Engineering
• Analytical Chemistry
• Endocrinology

Background:

• Type 2 diabetes mellitus (T2DM) management relies on frequent blood glucose monitoring.
• Current methods are invasive, potentially leading to poor patient compliance.
• Non-invasive biomarkers are sought to improve diabetes care.

Purpose of the Study:

• To investigate volatile organic compounds (VOCs) in breath as non-invasive biomarkers for blood glucose levels in T2DM patients.
• To explore the correlation between breath VOCs and blood glucose readings.
• To assess the feasibility of at-home breath sensor data collection for clinical studies.

Main Methods:

• A pilot clinical study involving six T2DM patients.
• Participants collected breath data using a custom sensor device at home.
• Breath data were correlated with finger-stick blood glucose readings using machine learning models (Support Vector Machine, Random Forest).

Main Results:

• Machine learning models achieved high accuracies: 85% (SVM) and 82% (Random Forest).
• Demonstrated feasibility of at-home breath sensor data collection for clinical research.
• Indicated potential for breath analysis as an alternative to invasive glucose monitoring.

Conclusions:

• Breath VOC analysis shows potential as a non-invasive method for glucose monitoring in T2DM.
• At-home breath sensor technology is viable for clinical studies.
• Further research with larger datasets and additional variables can enhance predictive power for improved diabetes management.