A wearable alcohol biosensor: Exploring the accuracy of transdermal drinking detection

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Machine Learning Applications

Background:

• Alcohol is detectable in sweat, but its transdermal emission is complex and varies with environmental conditions.
• Previous methods for real-time alcohol detection were limited by environmental factors and lag times.
• The feasibility of accurate, real-time drinking detection in diverse settings remained unclear.

Purpose of the Study:

• To evaluate the accuracy of a new generation of rapid-sampling transdermal biosensors for contemporaneous drinking detection.
• To explore the sensor's performance across diverse environmental conditions using machine learning.
• To assess the feasibility of real-time alcohol consumption monitoring.

Main Methods:

• 100 regular drinkers participated in laboratory sessions with manipulated alcohol doses and environmental conditions.
• Participants wore wrist sensors sampling at 20-second intervals, collecting transdermal alcohol data.
• Machine learning models translated transdermal data into alcohol use estimates, integrating data prior to breath alcohol concentration (BAC) assessment.

Main Results:

• Over 5.39 million transdermal readings and 12,699 BAC readings were collected.
• Transdermal sensors demonstrated strong accuracy for real-time drinking detection (AUROC 0.966, Sensitivity 89.8%, Specificity 90.6%).
• Models accurately differentiated high-risk drinking levels (AUROC up to 0.941).

Conclusions:

• Transdermal alcohol sensors show high accuracy for real-time drinking detection in both lab and field settings.
• The technology holds significant potential for long-term health tracking and medical monitoring.
• These sensors could be valuable for just-in-time relapse prevention strategies.