Full CGM + Meal Log Analysis
Feb 19 - Mar 17, 2026
Data
- CGM: 7,333 readings, Feb 19 22:21 to Mar 17 12:35
- Meal log: 52 meals, Feb 28 to Mar 16 (17 days with overlap)
Period Comparison
| Period | Mean | CV | TIR | >140 | >160 | Overnight (4-6am) |
|---|---|---|---|---|---|---|
| Pre-Lent (Feb 19-23) | 120 | 11.7% | 92% | 8.2% | 2.0% | 123 |
| Early Lent OMAD (Feb 24 - Mar 2) | 117 | 11.5% | 95% | 5.4% | 1.5% | 119 |
| Mid Lent (Mar 3-9) | 117 | 13.2% | 93% | 7.0% | 1.8% | 117 |
| Late Lent (Mar 10-17) | 114 | 9.9% | 97% | 2.5% | 0.5% | 114 |
Weekly Progression
- Week 8: mean 119, TIR 94%
- Week 9: mean 119, TIR 93%
- Week 10: mean 117, TIR 94%
- Week 11: mean 115, TIR 97%
- Week 12: mean 109, TIR 100%
Food Category Responses (avg delta / avg peak)
- Rice: +84 / 188 (n=1) — WORST
- Fruit/sweet: +31 / 144 (n=10)
- Pasta: +28 / 134 (n=2)
- Bread/toast/pastry: +21 / 132 (n=12)
- Protein-heavy: +18 / 130 (n=7)
- Greek yogurt/protein: +15 / 126 (n=7)
- Mixed/balanced: +13 / 129 (n=11)
- Eggs: +12 / 129 (n=2)
Fasting Duration vs Pre-Meal Glucose
Correlation: r = -0.457 (moderate negative) Longer fasts do correlate with lower pre-meal BG, but not dramatically.
Reactive Hypoglycemia
16 events total. Mostly Feb 20 and scattered through March. Pattern: big carb spikes (>140) followed by drops to 72-89.
Key Findings
- WHAT > WHEN: food composition matters more than eating window
- Rice is kryptonite (+84)
- Overnight glucose trending down: 123 → 114 over 4 weeks
- Weekly means dropping: 119 → 109
- Mar 16 was best day: mean 108, max 126, TIR 100%
- Protein + fat meals barely register on CGM
- Reactive hypo happens after big spikes — insulin overshoot
Exercise + CGM Cross-Reference (added Mar 17)
Activities in CGM window (Feb 19 - Mar 17):
- 7 VirtualRide sessions (29-60 min Z2)
- 1 WeightTraining (21 min KB complex)
- 1 Walk (20 min)
Exercise days vs rest days:
- Exercise days (n=7): mean 118.6, TIR 92.5%
- Rest days (n=20): mean 117.2, TIR 92.0%
- NOT significantly different overall (confounded by Mar 7 pancake disaster on exercise day)
Overnight glucose after exercise vs rest:
- After exercise: 116.8
- After rest: 118.0
- Marginally better but not dramatic
GLUT4 window (post-exercise meal spike by time gap):
- 0-1h: avg delta +20 (n=2)
- 1-2h: +40 (n=1, burritos)
- 2-4h: +28 (n=4)
- 4-8h: +28 (n=6)
- 8h+: -5 (n=1, steak)
- Rest day carb meals: avg delta +42 (n=10)
Reactive hypoglycemia:
- 10 events total (BG <90 within 6h of spike >125)
- 7 on exercise days, 3 on rest days
- Avg peak before: 150, avg nadir: 84, avg drop: 66, avg time: 2.8h
- Exercise doesn’t CAUSE hypo but amplifies it: depleted glycogen + carb spike → bigger insulin overshoot relative to available glucose
- Higher calorie burn → deeper hypo for same carb load (Feb 20: 197cal→drop 57; Mar 9: 481cal→drop 67)
- Exception: Mar 14 (449cal) had only 37-point drop because peak was only 126 (protein-dominant eating)
Key finding: Exercise + protein-dominant eating = best combo (Mar 14 template). Exercise + carb-heavy eating = worse hypo risk (Mar 9 spaghetti crash, Feb 20 nicnac crash).
Kelly Allocation Model (Mar 17, 2026)
Estimated potency coefficients:
- a (exercise): 0.42 mg/dL improvement per hour of Z2
- b (GNG): 0.57 mg/dL improvement per hour of GNG
- Ratio a/b: 0.73
- Optimal allocation: ~41% exercise, ~59% GNG
Key finding: The model works mathematically but food quality dominates both exercise and fasting effects by a wide margin. Time trend (r=-0.49) was the strongest predictor of improvement, meaning steady cumulative adaptation matters more than daily allocation.
Simulation projections (4 weeks from overnight 114):
- Max exercise: 114 (no change)
- Max fasting: 110 (-4)
- Current protocol: 113 (-1)
- Mar 14 template (1h Z2 + protein-dominant): 108 (-6) ← BEST
Conclusion: Kelly model confirms ~60/40 fasting-biased allocation, but the optimization target should be food quality on eating days, not hours of exercise vs fasting. One carb-heavy meal erases 5+ hours of either therapy.
Unified Glycogen Depletion Model (Mar 17, 2026)
Hypothesis: Exercise and fasting are the same therapy (glycogen depletion), just at different rates. Total “therapeutic hours” (liver glycogen <30g) should predict outcomes.
Result: MODEL FAILS
- Therapeutic hours vs same-day mean: r = -0.07 (essentially zero)
- Depletion integral vs same-day mean: r = -0.03
- Time trend vs overnight BG: r = -0.56 (still strongest)
Why it fails:
- Fasting days have max depletion (24h therapeutic) but mean BG 106-123 (GNG raises glucose)
- Eating days with protein-dominant meals have LOW depletion but best means (108-112)
- Mar 3: 24h deep depletion, integral 703 → mean 119
- Mar 14: 11h therapeutic, integral 104 → mean 112
Conclusion: Exercise and fasting are NOT interchangeable therapies:
- Fasting = hepatic fat oxidation → systemic adaptation over weeks
- Exercise = GLUT4 + glycogen cycling → acute disposal
- Food quality = prevents spikes → dominant same-day effect
Gamification score also failed: Best CGM days (Mar 14, Mar 16) scored in bottom 5 of depletion score. Need to rebuild around weekly trend protection, not daily depletion maximization.