Loading
Eight inputs, eight weights, one sum. The arithmetic is fully exposed, because the arithmetic is the only thing here worth trusting.
The figure below is a teaching illustration of how composite risk scores are built. It is not a clinical measurement, it has never been validated against mortality data, and it is not a substitute for medical care. A real biological-age estimate comes from an assay, fitted to a cohort, with a published error bar.
With no optional inputs answered the offset is zero, and the estimate is just your chronological age restated. Fill in a field to see a term appear.
| Input | Your value | Reference | Contribution |
|---|---|---|---|
| Resting heart rate | not provided | 65 bpm | — |
| Weekly activity | not provided | 150 min/wk | — |
| Smoking status | not provided | Never smoked | — |
| Average sleep | not provided | 7.5 h/night | — |
| Body-mass index | not provided | 20–25 | — |
| Systolic pressure | not provided | 115 mmHg | — |
| Self-rated health | not provided | Good (3 of 5) | — |
| Total offset | Sum of 0 answered terms, clamped to ±15 years | ±0.00 yr | |
Positive contributions push the estimate above your chronological age; negative ones pull it below. Terms you left blank are excluded from the sum rather than filled with the reference value, so a mostly-empty form produces a mostly-empty answer.
Each input is compared with a reference value. The difference is multiplied by a fixed weight to produce a contribution in years, positive or negative. The contributions are added. The sum is the offset, and the offset is added to your chronological age. That is the entire model. There is no training set, no regression, no held-out cohort.
The weights were chosen by hand to match the direction and the rough relative size of associations that replicate across large cohort studies: smoking dominates, self-rated health carries surprising weight for a single subjective question, activity and resting heart rate matter in the same ballpark as each other, and blood pressure scales roughly linearly above about 115 mmHg. They were not chosen to reproduce any published clock, and they would not survive contact with one.
Seven independent terms, each capped individually, can still sum to something absurd if a reader enters extreme values in every field. Real composite scores handle this with interaction terms and shrinkage fitted to data. This one has neither, so it clamps instead, and says so when the clamp bites. An offset of ±15 years is already at the far edge of what any questionnaire-based instrument claims.
An epigenetic clock reads DNA methylation at a few hundred to a few hundred thousand CpG sites and feeds them into a model trained on a cohort with known ages or known mortality outcomes. Second-generation clocks such as PhenoAge and GrimAge are trained on outcomes rather than chronological age, which is why they predict mortality better and correlate with chronological age worse. All of them report an error band. All of them disagree with each other on individual samples. The literature on whether an intervention can move a clock in a way that means anything is still open.
If a number here worries you, that is a reason to talk to a clinician, not a reason to trust the number.