Cardiorespiratory fitness — VO₂ max — is the strongest modifiable predictor of cardiometabolic death. Every major wearable already estimates it. No one manages it as a clinical signal. We do.
A panel of labs — HbA1c, blood pressure, BNP, BMI — measures single organs. VO₂ max measures the integrated output of the whole system — heart, lungs, blood, vessels, muscle metabolism. In head-to-head cohorts it equals or beats those same risk factors at predicting death.
And it is the only one estimable passively and continuously from a wearable already on hundreds of millions of wrists.
All-cause mortality falls when patients move from the unfit to the fit quartile.
Mandsager, JAMA Netw Open 2018
Reduction in cardiometabolic mortality per unit VO₂ improvement, with no observed ceiling.
Kodama, JAMA 2009
One in four US healthcare dollars. AHA projects 61% of adults will have CVD by 2050.
AHA · ADA 2024
Interval training raises VO₂ max with a large effect size while improving glucose, BP, and body composition.
Systematic review · 2024
The lowest fitness quintile carries a hazard ratio greater than smoking, diabetes, and coronary artery disease combined.
Source · Mandsager et al., JAMA Network Open, 2018
Percentage increase in all-cause mortality relative to the reference group. VO₂ max dwarfs every other modifiable risk factor.
Sources · Mandsager 2018 · Blair 1989 · Kodama 2009 · Framingham
“Mounting evidence over the past three decades has firmly established that low levels of cardiorespiratory fitness are associated with a high risk of cardiovascular disease and all-cause mortality… CRF is currently the only major risk factor not routinely assessed in clinical practice.”
560M+ devices output a VO₂ max estimate passively — Apple Watch, Garmin, Whoop, Fitbit, Polar, Oura.
The American Heart Association already designates fitness a clinical vital sign. The tooling has not caught up.
State-space filtering plus wearable foundation models convert noisy free-living data into a calibrated trajectory. Fenland: r ≈ 0.82.
A trivial two-input estimate — age and resting heart rate — already places most people inside a Mandsager quintile with a hazard ratio attached. This is what our platform tracks continuously, at higher fidelity, and closes the loop on.
Formula: VO₂ ≈ 15 × (HRmax / HRrest) · Uth-Sørensen-Overgaard-Pedersen
HRmax = 208 − 0.7 × age · Tanaka, JACC 2001
Norms: FRIEND registry · Kaminsky, Mayo Clin Proc 2015
Good · age/sex adjusted
vs elite baseline · Mandsager 2018
Estimate only. For clinical precision request a CPET (cardiopulmonary exercise test). Smartwatch VO₂ max readings are also a valid input.
A minimal personal ledger. Each 1 ml·kg⁻¹·min⁻¹ gain maps to ~13% lower all-cause mortality and ~45 additional life-days. Stored in your browser — the same longitudinal shape our platform manages at population scale.
Stored locally in your browser only. Export/import to move history. Forecast uses the Sparse Laplace Approximation Method (SLAM; Tillinghast, arXiv:1504.06352) with a log variance-stabilizing transform and a 30-day half-life kernel. Trend math uses Kodama 2009 and the Copenhagen 46-year cohort. See methods.
Log two or more readings and your trajectory appears here.
Fuse repeated free-living readings into a filtered latent state, overcoming the wide per-reading error of any single wearable.
Kalman / EKF / UKF tracking, hidden-Markov segmentation, particle filters, and Gaussian processes recover the true trajectory with calibrated uncertainty.
LSTM and transformer encoders, DeepSurv / DeepHit survival models for time-to-event risk, and reinforcement-learning dosing of intervention.
Incumbents optimize adjacent signals — weight, glucose, consumer fitness scores. None operationalize continuously-estimated fitness as a clinical management signal.
| Capability | Wearable OEMs | Twin · Virta · Omada | CGM platforms | AeroGlyphics |
|---|---|---|---|---|
| VO₂ max as clinical signal | Consumer score | Not tracked | Not tracked | Managed vital sign |
| Estimation strategy | Single-read, wide error | N/A | N/A | Filtered latent trajectory |
| Intervention loop | None | Manual coaching | Glucose-only | RL-dosed interval Rx |
| Reimbursement path | None | Employer / cash | Category-specific | Medicare RPM · SaMD |
| Hardware dependency | OEM lock-in | Devices + coaches | Sensor lock-in | Hardware-agnostic |
Cardiovascular digital health grows from ~$42B (2024) to ~$141B (2030E) at a 22.5% CAGR. Medicare RPM codes already reimburse device-based physiologic monitoring at $120+/patient/mo. The SaMD regulatory path is defined. The AHA vital-sign mandate precedes the tooling.
Benchmark filtered wearable trajectories against gold-standard cardiopulmonary exercise testing across a mixed cohort.
Head-to-head vs Framingham, PCE, and lab panels on linked EHR outcomes data — hazard for CV events and all-cause mortality.
Randomized: our estimation + dosing loop vs usual care. Endpoints: VO₂ change and measured cardiometabolic risk at 26 weeks.
We ride the wearable distribution instead of competing with it. 560M devices become our sensor fleet on day one.
Every filtered trajectory trains our proprietary risk and dosing models. Advantage compounds with cohort size.
A clinical vital sign, not a consumer score. Medicare RPM and SaMD paths, not app-store TAM.
Guideline mandate, wearable ubiquity, and model maturity all arrived at once. This window did not exist in 2022.
No proprietary claims dressed as science. Every stat we quote traces to published, replicated cardiology literature — the same evidence that pushed the AHA to designate fitness a clinical vital sign.
Mandsager K, Harb S, Cremer P, et al.
n = 122,007. Lowest fitness quintile carried 5.04× mortality vs elite. No upper limit of benefit.
Ross R, Blair SN, Arena R, et al.
CRF is currently the only major risk factor not routinely assessed in clinical practice.
Kodama S, Saito K, Tanaka S, et al.
Each 1-MET gain reduced all-cause mortality 13% and cardiovascular events 15% across 33 studies.
Blair SN, Kohl HW III, Paffenbarger RS Jr, et al.
First large prospective cohort establishing fitness as an independent mortality predictor.
Kokkinos P, Faselis C, Samuel IBH, et al.
US Veterans (n > 750,000). Modest fitness gains cut mortality 13–15%, independent of age, BMI, sex.
Brage S, Westgate K, et al.
Wearable-derived VO₂ max correlated with lab measurement at r ≈ 0.82 across 12,000+ adults.
Kaminsky LA, Arena R, Myers J.
First US CPET reference standards. Age- and sex-specific percentiles from > 7,700 verified maximal tests, now the FRIEND normative baseline.
Contemporary review
Positions VO₂ max as central to risk stratification in HF, CAD, valvular and pulmonary disease; identifies wearable estimation and closed-loop programming as the two largest unmet gaps.
Aeroglyphics is applying to Y Combinator, Techstars, and Stanford StartX. The full data room — thesis deck, scientific appendix, and program protocol — is available on request.