Methodology

1. The model

Each marker is converted into a matched age: the age at which your value would be typical for your sex, or the number of years that a deviation from a clinical target is worth. Your biological age is the weighted mean of those matched ages, plus a bounded lifestyle modifier for physical activity and sleep. The result is clamped to ±15 years around your chronological age and to the 18–80 range supported by the reference data.

matchedAge_i = clamp(age + z_i · dir_i · yearsPerSD_i,  age ± 15,  [18, 80])
      z_i = (value_i − μ_i(age, sex)) / σ_i(age, sex)

optimalBand   deviation = max(0, lo − value) + max(0, value − hi)
              matchedAge_i = clamp(age + (deviation / σ_i) · yearsPerSD_i,  age,  age + 15)

optimalFloor  matchedAge_i = age + max(0, threshold − value) / σ_i · yearsPerSD_i

BA_fit = Σ(w_i · matchedAge_i) / Σ w_i
BA     = clamp(clamp(BA_fit + lifestyle,  age ± 15),  [18, 80])
delta  = round1(BA − age)          Σ factor.impact === delta

2. Marker directions

Not every marker gets healthier in a straight line, and population averages are not health targets. Anchoring everything to the population mean would grade you on a sick curve — so markers whose average is distorted by age or treatment are anchored to clinical thresholds instead.

higherWorse

Higher is worse — the value is compared with the population mean for your age and sex.

higherBetter

Higher is better — same comparison, direction inverted.

optimalBand

Optimal band — deviation from a fixed clinical range in either direction ages you; inside the range is neutral and never makes you younger.

optimalFloor

Optimal floor — only values below a clinical threshold age you; at or above it is neutral.

3. Markers, weights and references

Weights reflect how informative each marker is about biological aging and are normalised across whatever you actually provide. Years per SD is the calibration coefficient: one standard deviation of deviation is worth this many years.

4. Attribution

Each factor card shows how many years that marker contributed. Contributions are distributed with largest-remainder rounding, so they always add up exactly to your total difference — no hidden residual.

5. Biological age vs optimal age

Biological age tells you where you stand relative to your peers. Optimal age tells you where you could stand: it recomputes the same weighted mean after neutralising every aging marker while keeping the advantages you already have. The gap between the two is the number of years the model considers recoverable.

6. Data sources

Population means, standard deviations and percentiles come from NHANES 2017–2018 (National Health and Nutrition Examination Survey, US CDC), restricted to adults aged 18–80 who completed both the interview and the physical examination, stratified into six age bands by sex. Values between band midpoints are linearly interpolated; outside the 24–75 range the nearest endpoint is used without extrapolation. Physical activity and sleep reference values come from published CDC estimates.

Age-band midpoints: 18-29 → 24 · 30-39 → 34.5 · 40-49 → 44.5 · 50-59 → 54.5 · 60-69 → 64.5 · 70-80 → 75

The dataset is rebuilt from the raw CDC files by scripts/process_nhanes.py and shipped as a static JSON file.

• NHANES 2017–2018, US Centers for Disease Control and Prevention
• WHO — Body mass index classification
• Whelton et al. — 2017 ACC/AHA Hypertension Clinical Practice Guideline
• Grundy et al. — 2018 ACC/AHA Cholesterol Clinical Practice Guideline
• Ashwell et al. — Waist-to-height ratio meta-analysis, Obesity Reviews 2012 · NICE CG189
• Holick et al. — Endocrine Society Vitamin D Clinical Practice Guideline, 2011
• Levine et al. — PhenoAge, Aging 2018

7. Limitations

This is a statistical estimate, not an epigenetic clock. It is calibrated on a US population sample and may fit other populations less well. Self-reported activity and sleep are noisy. The lifestyle modifier is deliberately capped at ±3 years so that questionnaire answers cannot outweigh measured markers. Nothing here is a diagnosis — abnormal values should be discussed with a healthcare professional.