RatioLogo
Back

From Lab to Code: A New Era for Glycemic Index Testing

For decades, determining the Glycemic Index (GI) of a new food has been a resource-heavy ordeal. Following rigorous standards, it requires recruiting human volunteers for hours of manual blood monitoring. This process is slow and prone to the messy variability of human biology.

Now, a team of researchers has developed a mathematical "shortcut" that could move GI testing from the clinic to the computer. By using a complex system of equations, scientists have successfully reconstructed individualized glucose curves for a cohort of 35 healthy subjects.

The Traditional Human-Led Process

Following the ISO 26642:2010 standard, determining a food's GI is a manual, human-centric ordeal.

The method requires:

  • Recruiting at least ten healthy volunteers.
  • Subjecting them to strict 12-hour fasts.
  • Repeatedly drawing blood to monitor their biological response to 50g of pure glucose.

The New Algorithmic Approach

This discovery paves the way for the automatic computation of nutritional data. By mathematically simulating how a body processes sugar, we could skip the needles and provide highly specific dietary planning.

The core innovation uses a modified Dalla Man–Rizza–Cobelli maximal model. This is a complex system of ordinary differential equations that can reconstruct a person's glucose response curve digitally.

The Discovery of Metabolic Phenotypes

The data-driven analysis of the glucose standard identified three distinct metabolic phenotypes, categorized by how quickly blood sugar peaks after ingestion.

Group 1: The Early Peakers

  • Demonstrated a lightning-fast metabolic response.
  • Glucose absorption rate (kₐբₛ): 0.2917 ± 0.0062 min⁻¹.
  • Stomach "grinding" rate (k₉ᵣᵢ): 0.0939 ± 0.0033 min⁻¹.

Group 3: The Late Peakers

  • Processed sugar at a significantly slower rate.
  • Glucose absorption rate (kₐբₛ): 0.0861 ± 0.0136 min⁻¹.
  • Stomach "grinding" rate (k₉ᵣᵢ): 0.0436 ± 0.0086 min⁻¹.

Key Insight: The research found that two primary levers drive these differences:

  1. The intestinal absorption rate (kₐբₛ).
  2. The stomach's mechanical "grinding rate" (k₉ᵣᵢ).

These parameters allow the model to predict exactly how a person will handle a sugar spike.

Current Limits and the Path Forward

While the results suggest a future where food labels are generated by algorithms, the researchers acknowledge the model's current boundaries.

Important limitations include:

  • Demographic Focus: The study used a homogenous group of subjects aged 20–40. The math may not yet apply to pediatric or diabetic populations.
  • Dietary Simplicity: The model was tested only on 50g of pure glucose. Real-world meals contain fats, proteins, and fibers that slow digestion.
  • Future work is needed to see if the equations can handle the complexity of a real meal as easily as a glass of sugar water.

Reference: Credali, F., Venuti, M. T., Boffi, D., and Rossi, P. (2025). Automatic Computation of the Glycemic Index: Data Driven Analysis of the Glucose Standard. arXiv:2506.15471v1 [math.DS].