Science & methodology
Biology is complex. A useful summary should make that complexity easier to explore, while keeping its assumptions and uncertainty in view.
The reporting language
Δ-years is a BioMIR model output expressed in age-equivalent units relative to chronological age. Negative or positive Δ-years indicate position on the model’s reference scale; they are not rewards, penalties, or direct measurements of years gained or lost.
It does not measure years added to or removed from your life. A daily change does not establish irreversible aging, rejuvenation, or a change in lifespan.
The scientific architecture
BioMIR separates a daily physiological trajectory, an action-linked credit model, and episodic clinical comparators. They can be examined together without treating their outputs as interchangeable.
01 / Daily longitudinal model
Combines the Cardiometabolic Age baseline with Behavioral and Functional Δ-year contributions. Its domains and biomarker attribution explain the daily estimate and how its trajectory changes over time.
Explore the model specification →02 / Action-linked model
Accounts for modeled ΔYears credit earned from completed eligible actions against frozen opportunities. REC is a distinct model informed by ABA-ranked contributors; its credit is not an observed change in ABA or physiological age.
Explore the REC model →03 / Laboratory-based models
KDM and clinical Levine PhenoAge are computed separately from eligible dated laboratory panels. Their episodic readouts help examine longer-term directional concordance with the daily trajectory.
Examine clinical methodology →Inside Adaptive BioAge
Adaptive BioAge brings together Behavioral, Functional, and Cardiometabolic contributions. Keeping the domains visible helps distinguish a recorded behavior from a physiological response and a more persistent pattern.
Behavioral
Steps, Dietary Energy, Dietary Carbohydrates, Time in Daylight, Alcohol, and Dietary Sodium are the active Behavioral inputs; Meditation is not active in scoring. A reported behavior can change immediately, while any physiological association may be delayed, nonlinear, bidirectional, or confounded.
Daily exposures
patterns across weeks
Functional
Heart rate variability, resting heart rate, VO₂max, Cardio Recovery, Walking Speed, and Total Sleep provide complementary context about autonomic regulation, cardiorespiratory capacity, recovery, and mobility. Some readings vary day to day; fitness adaptation generally requires a longer view.
Daily variation
with longer-term context
Cardiometabolic
Systolic blood pressure, the selected glucose observation, and body mass index provide the slower-moving Cardiometabolic foundation for the daily model. Individual readings can vary, but their meaning is strongest when interpreted as repeated observations and trajectory over time. Together, they provide cardiometabolic context alongside faster-changing Functional and Behavioral signals.
Repeated observations
over weeks and months
These are interpretation guides, not fixed biological deadlines. Measurement conditions and the specific input matter.
Under the model
Cardiometabolic Age uses age-regressed population coefficients for systolic blood pressure, the selected glucose observation, and BMI to construct an age-equivalent baseline. This layer is distinct from the bounded Behavioral and Functional modifiers and from the separate KDM clinical clock.
Behavioral and Functional inputs use standardized comparisons with model-specified reference values, including sex-stratified and age-band context where applicable. They are not themselves age-regression clocks. Their domain effects are translated into the app’s common age-equivalent Δ-years reporting language only after reference normalization, direction alignment, and weighting.
Behavioral and Functional inputs contribute according to published evidence and model-specific design decisions, rather than assuming equal importance for every input. Domain constraints limit the aggregate influence of more variable inputs, with intentionally asymmetric handling of favorable and adverse contributions. These are modeling choices requiring independent evaluation. Epidemiologic associations are not the causal effect, absolute risk, or clinical prognosis of an individual. Read the public framework and interpretation boundaries.
Data Confidence summarizes support for the daily estimate, while Data Freshness preserves the distinction between a retained value and a newly observed one. These indicators describe support and recency—not device accuracy, statistical certainty, or probability that a model result is correct. Methods specifies the stream-level support and carry-forward rules.
The evidence-to-action system
Reclaimed BioAge (REC) is BioMIR’s distinct action-linked biological-age model. ABA-ranked biomarker contributors inform eligible action opportunities; REC then accounts for modeled age-equivalent credit earned through recorded completion. The eligible ΔYears denominator is frozen for the reporting period. Skipped, unresolved, and eligible uncompleted opportunities remain in that denominator and earn zero credit.
Two percentages answer different questions. Completed Actions (%) counts completed unique eligible actions against reportable unique eligible actions. Reclaimed ΔYears (%) is earned modeled credit divided by frozen eligible ΔYears opportunity. Credit cannot exceed its linked opportunity. REC therefore quantifies modeled follow-through, not physiological change; it is interpreted beside, rather than added to or subtracted from, ABA and the clinical clocks. See accounting rules in Methods.
Model contributors help order eligible signals for the action list. Priority indicates what the model highlights; it is not a diagnosis or a clinical risk ranking.
Available biomarker context informs a graduated action option. When several signals point to the same commitment, the list brings them together so the person sees one coherent action.
Each eligible opportunity retains its linked modeled ΔYears allocation. Shared commitments and unresolved outcomes require consistent accounting so a period’s numerator and denominator describe the same eligible record set.
Inspect completed-action percentage and opportunity-weighted reclaimed credit across the selected period, then compare the record with independently observed Functional, Cardiometabolic and clinical biomarker trajectories. Temporal agreement is a research question, not a built-in causal conclusion.
Explore the evidence context · See how Today Actions work in the app · Review validation status
REC closes the evidence-to-action loop without converting action completion into measured biological benefit. Independent evaluation can ask whether sustained recorded follow-through anticipates favorable subsequent biomarker trajectories. Review the three REC research workstreams.
A separate clinical-panel perspective
These established but methodologically distinct models provide age-equivalent context from eligible, episodic laboratory panels; neither is mathematically incorporated into ABA or REC. The observation date, panel eligibility, model version and Data Freshness matter when interpreting a sequence of clinical values. Clinical and daily estimates may show directional concordance without equal numerical values or synchronized response times. Methods explains carry-forward and display conventions.
KDM estimates an age-equivalent physiological profile from relationships among age, a biomarker panel, and biomarker-specific variation. It is a statistical age model—not a prognosis or a direct measurement of remaining lifespan.
Read the original methodLevine PhenoAge combines chronological age with nine laboratory biomarkers using an outcome-linked model. It is distinct from DNAm PhenoAge, the related DNA methylation model.
Read the clinical Phenotypic Age researchEvidence for an established method does not automatically validate BioMIR’s implementation, daily composite, classifications, or changes in Δ-years. Those require separate evaluation.
Read the evidence in layers
Studies examine associations, mechanisms, and interventions under specific conditions. Their populations, endpoints, and study designs determine what can be concluded.
Reference selection, weighting, transformations, and aggregation shape the output. Scientific plausibility and a transparent design are valuable; external validation is a separate task.
N=1 observations can reveal repeatable personal patterns. They do not, by themselves, prove causation. Changes in devices, coverage, routines, and measurement timing can also explain a trend.
Compare the same input, time window, and statistic. Check Data Confidence, whether cardiometabolic values are current or carried forward, and the original measurement date. Look for a sustained pattern before attributing a result to an action.
Continue exploring
Read the implementation rules in Methods, then trace the supporting literature in Resources.