Researchers unveil digital 'alcohol twin' to predict drinking risks and guide personalized interventions

By Vijay Kumar MalesuReviewed by Benedette Cuffari, M.Sc.May 9 2024

In a recent study published in the journal npj Digital Medicine, researchers developed a digital twin model that simulates real-life alcohol consumption and links it to long-term clinical biomarkers, thereby enhancing eHealth strategies to reduce dangerous drinking habits.

Study: A physiologically-based digital twin for alcohol consumption—predicting real-life drinking responses and long-term plasma PEth. Image Credit: niksdope / Shutterstock.com

Approaches to reduce alcohol abuse

Alcohol consumption accounts for about 5% of global deaths and can increase the risk of significant health issues like liver diseases and cancers. Chronic and binge drinking, both of which are frequent habits among young adults, can cause both immediate injuries and long-term health problems.

EHealth applications, including tools like estimated blood alcohol concentration (eBAC) calculators, have shown promise in reducing excessive drinking. However, models like the Widmark equation, which estimates blood alcohol levels, fail to capture the complexities of real-life drinking patterns, including interactions with different drink types and food.

Thus, an accurate measurement of alcohol consumption using markers like phosphatidylethanol (PEth) is essential for better interventions. However, additional research is needed to enhance these predictive models and link them to health outcomes.

About the study 

In the current study, the digital twin model employs a series of equations to describe the physiological processes involved in alcohol use. Initially, the model leverages ordinary differential equations (ODEs), illustrating how the state variable changes in response to reaction rates and inputs.

The model intricately details the dynamics of gastric emptying by incorporating variables like drink volume and caloric content, both of which affect the stomach's volume and rate of emptying. These dynamics are influenced by the caloric content of liquids and the presence of solid food, with specific equations modeling how calories from food slow gastric emptying.

The model also addresses ethanol metabolism, detailing how ethanol interacts with meals in the stomach. More specifically, ethanol is temporarily encapsulated within food, thereby modifying its availability for absorption. The model describes this interaction, as well as subsequent ethanol release and metabolism through enzymatic pathways in the liver, such as alcohol dehydrogenase (ADH) and cytochrome P450 2E1 (CYP2E1), which also produce metabolites like acetate and PEth.

The initial conditions assume the individual begins in a fasted state with no residual ethanol. The model's parameters are carefully defined, and optimal values are estimated from empirical data to ensure their accuracy and reliability.

a Short overview of the physiological process that the model describes. Ethanol enters the body via the stomach, where already a small amount can enter plasma. Via the stomach emptying, the ethanol enters the intestine. Here, most of the ethanol is taken up via absorption. Most of the ethanol is metabolized in the liver, and a small amount is excreted via renal pathways. In the liver, ethanol is converted into acetaldehyde via three oxidative pathways governed by the enzymes: Alcohol dehydrogenases (ADH), catalase, and cytochrome P450 2E1 (CYP2E1). Acetaldehyde is further converted into acetate and then acetyl-COA. There also exist non-oxidative pathways, responsible for a miniscule amount of ethanol breakdown, e.g., into phosphatidylethanol (PEth). Following, the blood alcohol concentration (BAC), or the breath alcohol concentration (BrAC), is measured. These physiological processes can be described using a mathematical model, a physiologically-based digital twin. The digital twin can be used for several use cases, such as for education and awareness, in self-reporting and monitoring of alcohol consumption, and as a tool to support the combination of AUDIT and PEth reports. b Schematic over the modeling approach. c Schematic showing the model structure.

Study findings 

The physiologically-based digital twin model was rigorously trained and validated using a diverse array of published experimental data. This model framework successfully aligns with all data from the estimation dataset using consistent parameters. Furthermore, it accurately predicts outcomes from independent validation data, confirmed by χ²-tests under a 0.05 confidence level.

This novel mechanistic model excels in representing gastric emptying dynamics across different experimental conditions. Notably, previous studies have demonstrated that while the caloric content significantly impacts gastric emptying rates, the type of calories does not.

The model captures this phenomenon by depicting consistent emptying rates across different caloric types but varied rates based on total caloric content. This gastric emptying behavior, irrespective of caloric density, is effectively modeled, thus offering a detailed understanding of the involved physiological processes.

While exploring the interaction between meals and plasma ethanol dynamics, the model evaluated four hypotheses concerning how food affects ethanol metabolism. Of these, only the hypothesis that food encapsulates alcohol and releases it as the food is digested could adequately match the empirical data. This observation reflects a refined understanding of meal-induced modifications to ethanol absorption and metabolism.

The scope of the model extends to detailed scenarios of ethanol dynamics in the plasma following alcohol consumption with or without food. Moreover, it accurately reflects the impact of different alcoholic beverages on BAC by capturing variations in ethanol levels introduced by the presence of food. This includes data from studies showing how meals can modulate the peak and progression of BAC.

The model also describes the metabolic pathways of ethanol, including both oxidative and non-oxidative processes. These results were similar to previous experimental data on plasma acetate and PEth, both of which are important markers for studying alcohol consumption effects.

This predictive capability extends to personalized scenarios where the model considers individual anthropometric data to forecast the effects of different drinking patterns on BAC and PEth levels. By varying parameters such as body mass index (BMI), the model offers tailored insights into how specific drinking habits impact individuals differently, thereby enhancing its application in personalized medicine and public health strategies.

Conclusions

The model developed in the current study reflects advanced simulations in alcohol research and serves as a pivotal tool for future developments in eHealth applications aimed at managing and understanding alcohol consumption. Its ability to integrate complex biological interactions into a coherent framework makes it a valuable resource for clinicians and researchers interested in reducing alcohol-related harms through targeted interventions and personalized approaches.