The Prediction Model

Core Prediction Framework

Our glucose prediction model integrates four fundamental biomarkers to create personalized metabolic predictions with near clinical-grade accuracy.

📈

Glycemic Index (GI)

Speed of glucose absorption from 0-100 scale

High GI (≥70): Fast absorption, sharp peaks
Low GI (≤55): Slow absorption, gentle curves

⚡

Glycemic Load (GL)

Total metabolic impact = (GI × Carbs) ÷ 100

Low GL (<10): Minimal impact
High GL (≥20): Significant glucose response

🔬

Insulin Sensitivity

How effectively cells respond to insulin signals

Normal: Efficient glucose clearance
Resistant: Elevated, prolonged responses

🏃

Gastric Emptying

Rate of food transit from stomach to intestine

Fast: Earlier peaks, rapid absorption
Delayed: Later peaks, extended elevation

Metabolic State Profiles

✅

Normal Metabolism

Baseline: 85-90 mg/dL

Insulin Sensitivity: 100% (Optimal)

Peak Timing: Normal (GI-dependent)

Glucose Clearance: 60% decline/hour

Crash Risk: Minimal (high-GI only)

Efficient insulin response with gentle return to baseline

⚠️

Prediabetic

Baseline: 105-110 mg/dL

Insulin Sensitivity: 60-70% (Reduced)

Peak Timing: 20% delayed

Glucose Clearance: 25% decline/hour (fast crash)

Crash Risk: High (reactive hypoglycemia)

Insulin overcompensation leading to below-baseline crashes

🔴

Type 2 Diabetic

Baseline: 140-160 mg/dL

Insulin Sensitivity: 30-40% (Severely reduced)

Peak Timing: 40% delayed

Glucose Clearance: 35% decline/hour (slow)

Crash Risk: Low (stays elevated)

Prolonged elevation with gradual decline, rarely below baseline

Insulin Sensitivity Assessment

🏆 GOLD STANDARD

Hyperinsulinemic-Euglycemic Clamp (HE-Clamp)

The definitive clinical method for measuring insulin sensitivity with 95%+ accuracy.

Procedure:

1 Continuous insulin infusion to raise plasma insulin to ~100 μU/mL

2 Simultaneous glucose infusion to maintain euglycemia (90 mg/dL)

3 Measure glucose infusion rate (M-value) required for stability

M-Value Ranges:

Normal: >7.5 mg/kg/min
Insulin Resistant: 4-7.5 mg/kg/min
Severely Resistant: <4 mg/kg/min

Clinical Limitations:

Requires 3-6 hours in clinical setting
Expensive ($800-2000 per test)
Invasive (multiple IV lines)
Not practical for routine monitoring

🏠 AT-HOME SOLUTIONS

Accessible Insulin Sensitivity Calculations

We provide clinically-validated, at-home alternatives that correlate strongly with HE-clamp results.

QUICKI (Quantitative Insulin Sensitivity Check Index)

QUICKI = 1 ÷ [log(I₀) + log(G₀)]

I₀ = Fasting insulin (μU/mL)
G₀ = Fasting glucose (mg/dL)

Interpretation:

Normal: >0.357
Insulin Resistant: 0.320-0.357
Diabetic: <0.320

HOMA-IR (Homeostatic Model Assessment)

HOMA-IR = [I₀ × G₀] ÷ 405

I₀ = Fasting insulin (μU/mL)
G₀ = Fasting glucose (mg/dL)

Interpretation:

Normal: <2.5
Insulin Resistant: 2.5-6.0
Severe Resistance: >6.0

Clinical Validation:

QUICKI correlation with M-value: r = 0.78 (p < 0.001)
HOMA-IR correlation with M-value: r = -0.73 (p < 0.001)
Required tests: Fasting glucose + insulin (simple blood draw)
Cost: INR 600-1000 vs INR 20000+ for HE-clamp

Personalization Integration

📋

Assessment Input

User provides either:

Clinical M-value from HE-clamp
Home QUICKI calculation
Home HOMA-IR calculation

→

⚙️

Sensitivity Calibration

Model adjusts:

Peak timing delays
Glucose clearance rates
Crash susceptibility
Baseline glucose levels

→

🎯

Personalized Prediction

Generates:

Individual glucose curves
Personalized timing
Metabolic risk assessment
Tailored recommendations

Personalization Impact Example

Standard Prediction

Food: Apple (GI: 36, GL: 6)

Peak: 75 minutes, 125 mg/dL

Return: Baseline by 120 minutes

Personalized (HOMA-IR: 4.2)

Food: Same apple

Peak: 95 minutes, 145 mg/dL

Return: Still elevated at 180 minutes

25% delayed peak, 35% longer elevation

Future Model Enhancements

🏃‍♂️

Gastric Emptying Personalization

Integration of individual gastric emptying rates through:

Gastroparesis Cardinal Symptom Index
Home ultrasonic assessment
Meal transit time calculations

🔄

Dynamic Range Modeling

Moving beyond fixed values to adaptive ranges:

Food-dependent sensitivity modulation
Circadian rhythm integration
Cumulative metabolic memory

📱

Continuous Monitoring

Real-time model refinement through:

CGM data integration
Machine learning adaptation
Personalized coefficient adjustment

The Prediction Engine