Hypothesis

Multimodal wearable biomarkers (CGM, HRV, sleep) combined with activity-derived resilience metrics improve prediction of biological age acceleration beyond activity alone, and this improvement is mediated by mitochondrial hormesis pathways that link glucose variability to autonomic balance.

Rationale

• Activity-based deep learning models (e.g., GeroSense) capture resilience via autocorrelation of heart-rate variability and predict mortality with accuracy rivaling blood-based clocks【1】【2】【3】.
• Consumer CGMs and HRV monitors lack validation for biological-age prediction, yet glucose fluctuations and autonomic tone are known to influence mitochondrial NAD+/SIRT1 activity, a core regulator of stress resistance【4】.
• Psychological burden of continuous tracking can attenuate intrinsic motivation, but limiting measurement to strategically chosen windows reduces anxiety while preserving trend information【5】【6】.

Mechanistic Insight

We hypothesize that intermittent spikes in glucose variability trigger a mild oxidative stress that activates Nrf2-dependent antioxidant responses and boosts SIRT1 deacetylase activity. Simultaneously, heightened parasympathetic HRV reflects vagal tone that enhances mitochondrial efficiency through acetylcholine-mediated reduction of reactive oxygen species. When both signals converge, the cell experiences a hormetic state that improves resilience, slowing the Gompertz-linked decline in recovery speed. Conversely, discordant patterns—high glucose variability coupled with low HRV—indicate maladaptive stress and predict accelerated biological age.

Testable Predictions

1. In a sample of n=30 adults wearing a CGM, HRV chest strap, and actigraphy for 14 days, a multivariate model that includes glucose-variability metrics (e.g., mean absolute glucose change per hour), HRV-derived parasympathetic index (RMSSD), and sleep efficiency will explain significantly more variance in weekly resilience scores (derived from GeroSense-style autocorrelation decay) than activity-only models (ΔR² ≥ 0.12, p < 0.01).
2. The interaction term (glucose variability × HRV parasympathetic index) will be negative, indicating that high glucose variability only predicts lower resilience when parasympathetic tone is low.
3. Participants who receive personalized feedback limiting CGM/HRV alerts to two 30-minute windows per day will report lower state-anxiety scores (STAI-6) and maintain comparable trend-data quality (intraclass correlation ≥ 0.8) compared with continuous-feedback groups.

Falsifiability

If the multimodal model fails to outperform activity-only predictors (ΔR² < 0.02) or the interaction term is non-significant, the hypothesis is refuted. Likewise, if restricted feedback does not reduce anxiety or degrades trend reliability, the proposed psychological mitigation strategy is invalid.

Experimental Design (n=1 Adaptive Protocol)

• Baseline week: Wear devices continuously, collect raw data, compute resilience via rolling autocorrelation of HRV.
• Intervention weeks (4): Randomize days to either "full-stream" (alerts every 5 min) or "batch-stream" (alerts only at 09:00 and 21:00).
• Washout: 3-day period with no feedback to assess carry-over.
• Outcome: Weekly resilience slope, glucose variability, HRV parasympathetic index, sleep efficiency, and STAI-6.
• Analysis: Mixed-effects model with condition as fixed effect, participant as random effect; test predictions above.

By linking mitochondrial hormesis to multimodal wearable signals and controlling measurement burden, this hypothesis offers a concrete, falsifiable path toward validated consumer-device biomarkers of aging.