Published foundation
The blood biomarker layer is grounded in the Levine et al. (2018) PhenoAge research model. Vector adapts that method for the units printed on Australian pathology reports and combines it with longitudinal context to produce a consistent, personalised view of biological age direction.
Layer 1: nine core pathology inputs
The current app uses Albumin, Creatinine, Glucose, hs-CRP, an internal lymphocyte percentage, MCV, RDW, ALP and WBC alongside chronological age. The biomarker library exposes absolute Lymphocytes; the app derives the internal percentage from absolute Lymphocytes divided by WBC.
CRP is converted for the logarithmic term from Australian mg/L context. Creatinine and ALP receive sex-specific z-score normalisation before entering the formula.
Australian report-unit adaptation
Vector adapts the published foundation for the units printed on Australian pathology reports and calibrates the resulting distribution against a synthetic population. This keeps report units and collection context aligned with the records the product is designed to follow.
The result is a Vector-specific estimate grounded in the published Levine foundation, adapted for Australian reports. The adaptation is a product implementation that provides a consistent view of direction across a record; it is not a re-derivation of the original model.
Population standardisation
A raw PhenoAge output reads implausibly young for healthy young adults, a known age bias in the published model. Vector standardises the displayed difference against an age- and sex-matched population so the number means the same thing at 30 as it does at 60. The underlying pills and component scores stay raw.
Layer 2: secondary modifiers
Where suitable companion data exists, ApoB, HbA1c, Total Testosterone, SHBG through the Free Androgen Index and Oestradiol can contribute signed secondary adjustments. The app weights the layer by how many eligible markers are present, so one result cannot dominate the estimate.
Layer 3: wearable context
Apple Health signals such as HRV, resting heart rate, VO₂ max, training volume and sleep form a separate everyday context layer. Missing wearable values carry no hidden penalty — they reduce available context rather than declaring a problem.
Missing data, stale data and confidence
The estimate requires suitable, correctly dated values. Missing or stale inputs are surfaced in the app’s confidence and availability context. Vector does not backfill an absent biomarker with a healthy assumption, because doing so would turn an absence of evidence into a good result.
How the estimate is displayed
The backend estimate is clamped to a maximum difference of ±15 years from chronological age, and the display constrains the visible age to a sensible range. These bounds keep the result readable and comparable as a record grows, with the supporting data context visible beside it.
What the estimate helps you see
- Compare a new estimate against the direction of your own reviewed record.
- See which pathology markers and Apple Health signals informed the result.
- Use the trend to choose which patterns or follow-up questions deserve attention.
- Treat it as a wellness signal, never as a diagnosis or a clinical substitute.
Sources
- Levine et al. (2018), PhenoAgeThe published research foundation for the blood biomarker layer.
- AACB / RCPA Common Reference IntervalsHarmonised Australian adult reference intervals.
- RCPA Manual: Pathology TestsLaboratory reference interval context.